Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,000 --> 00:00:01,190
I want you to picture something for a
2
00:00:01,190 --> 00:00:01,200
I want you to picture something for a
3
00:00:01,200 --> 00:00:01,630
I want you to picture something for a
second.
4
00:00:01,630 --> 00:00:01,640
second.
5
00:00:01,640 --> 00:00:04,670
second.
>> Okay. Um, imagine you're standing at the
6
00:00:04,670 --> 00:00:04,680
>> Okay. Um, imagine you're standing at the
7
00:00:04,680 --> 00:00:07,470
>> Okay. Um, imagine you're standing at the
top of this really bumpy, uneven hill
8
00:00:07,470 --> 00:00:07,480
top of this really bumpy, uneven hill
9
00:00:07,480 --> 00:00:10,030
top of this really bumpy, uneven hill
and you drop a marble. Okay. And as that
10
00:00:10,030 --> 00:00:10,040
and you drop a marble. Okay. And as that
11
00:00:10,040 --> 00:00:12,710
and you drop a marble. Okay. And as that
marble rolls down, it starts hitting,
12
00:00:12,710 --> 00:00:12,720
marble rolls down, it starts hitting,
13
00:00:12,720 --> 00:00:14,470
marble rolls down, it starts hitting,
you know, little ridges and it bounces
14
00:00:14,470 --> 00:00:14,480
you know, little ridges and it bounces
15
00:00:14,480 --> 00:00:16,190
you know, little ridges and it bounces
into small valleys branching off into
16
00:00:16,190 --> 00:00:16,200
into small valleys branching off into
17
00:00:16,200 --> 00:00:18,070
into small valleys branching off into
all these different paths until it
18
00:00:18,070 --> 00:00:18,080
all these different paths until it
19
00:00:18,080 --> 00:00:19,870
all these different paths until it
finally comes to a complete stop at the
20
00:00:19,870 --> 00:00:19,880
finally comes to a complete stop at the
21
00:00:19,880 --> 00:00:20,310
finally comes to a complete stop at the
bottom.
22
00:00:20,310 --> 00:00:20,320
bottom.
23
00:00:20,320 --> 00:00:22,070
bottom.
>> Yeah, it is such a classic image. I
24
00:00:22,070 --> 00:00:22,080
>> Yeah, it is such a classic image. I
25
00:00:22,080 --> 00:00:24,350
>> Yeah, it is such a classic image. I
mean, in developmental biology we refer
26
00:00:24,350 --> 00:00:24,360
mean, in developmental biology we refer
27
00:00:24,360 --> 00:00:26,310
mean, in developmental biology we refer
to that as Waddington's landscape.
28
00:00:26,310 --> 00:00:26,320
to that as Waddington's landscape.
29
00:00:26,320 --> 00:00:26,710
to that as Waddington's landscape.
>> Right, yeah.
30
00:00:26,710 --> 00:00:26,720
>> Right, yeah.
31
00:00:26,720 --> 00:00:29,270
>> Right, yeah.
>> It was uh proposed decades ago to help
32
00:00:29,270 --> 00:00:29,280
>> It was uh proposed decades ago to help
33
00:00:29,280 --> 00:00:31,790
>> It was uh proposed decades ago to help
visualize exactly how a cell makes
34
00:00:31,790 --> 00:00:31,800
visualize exactly how a cell makes
35
00:00:31,800 --> 00:00:32,630
visualize exactly how a cell makes
decisions.
36
00:00:32,630 --> 00:00:32,640
decisions.
37
00:00:32,640 --> 00:00:33,910
decisions.
>> Yeah, and instead of a marble, we're
38
00:00:33,910 --> 00:00:33,920
>> Yeah, and instead of a marble, we're
39
00:00:33,920 --> 00:00:35,870
>> Yeah, and instead of a marble, we're
talking about a single generic stem
40
00:00:35,870 --> 00:00:35,880
talking about a single generic stem
41
00:00:35,880 --> 00:00:36,270
talking about a single generic stem
cell.
42
00:00:36,270 --> 00:00:36,280
cell.
43
00:00:36,280 --> 00:00:37,190
cell.
>> Exactly.
44
00:00:37,190 --> 00:00:37,200
>> Exactly.
45
00:00:37,200 --> 00:00:38,950
>> Exactly.
>> As it rolls down this metaphorical
46
00:00:38,950 --> 00:00:38,960
>> As it rolls down this metaphorical
47
00:00:38,960 --> 00:00:42,070
>> As it rolls down this metaphorical
landscape, it's um it's losing its
48
00:00:42,070 --> 00:00:42,080
landscape, it's um it's losing its
49
00:00:42,080 --> 00:00:43,510
landscape, it's um it's losing its
limitless potential, right? It's
50
00:00:43,510 --> 00:00:43,520
limitless potential, right? It's
51
00:00:43,520 --> 00:00:45,230
limitless potential, right? It's
figuring out its final identity. Like,
52
00:00:45,230 --> 00:00:45,240
figuring out its final identity. Like,
53
00:00:45,240 --> 00:00:46,870
figuring out its final identity. Like,
will it slide into the valley that makes
54
00:00:46,870 --> 00:00:46,880
will it slide into the valley that makes
55
00:00:46,880 --> 00:00:48,110
will it slide into the valley that makes
it a red blood cell?
56
00:00:48,110 --> 00:00:48,120
it a red blood cell?
57
00:00:48,120 --> 00:00:50,230
it a red blood cell?
>> Mhm. Or maybe it bounces a different way
58
00:00:50,230 --> 00:00:50,240
>> Mhm. Or maybe it bounces a different way
59
00:00:50,240 --> 00:00:51,990
>> Mhm. Or maybe it bounces a different way
and, you know, becomes a skin cell or a
60
00:00:51,990 --> 00:00:52,000
and, you know, becomes a skin cell or a
61
00:00:52,000 --> 00:00:54,350
and, you know, becomes a skin cell or a
neuron. Right. And the holy grail for
62
00:00:54,350 --> 00:00:54,360
neuron. Right. And the holy grail for
63
00:00:54,360 --> 00:00:56,190
neuron. Right. And the holy grail for
biologists has really always been trying
64
00:00:56,190 --> 00:00:56,200
biologists has really always been trying
65
00:00:56,200 --> 00:00:58,790
biologists has really always been trying
to map out that exact landscape. We want
66
00:00:58,790 --> 00:00:58,800
to map out that exact landscape. We want
67
00:00:58,800 --> 00:01:01,150
to map out that exact landscape. We want
to understand the invisible forces that
68
00:01:01,150 --> 00:01:01,160
to understand the invisible forces that
69
00:01:01,160 --> 00:01:03,590
to understand the invisible forces that
push the marble down one specific valley
70
00:01:03,590 --> 00:01:03,600
push the marble down one specific valley
71
00:01:03,600 --> 00:01:05,950
push the marble down one specific valley
instead of another because
72
00:01:05,950 --> 00:01:05,960
instead of another because
73
00:01:05,960 --> 00:01:07,710
instead of another because
I mean, if we know that, we understand
74
00:01:07,710 --> 00:01:07,720
I mean, if we know that, we understand
75
00:01:07,720 --> 00:01:09,590
I mean, if we know that, we understand
how life builds itself from scratch.
76
00:01:09,590 --> 00:01:09,600
how life builds itself from scratch.
77
00:01:09,600 --> 00:01:11,990
how life builds itself from scratch.
>> Which brings us to the mission for
78
00:01:11,990 --> 00:01:12,000
>> Which brings us to the mission for
79
00:01:12,000 --> 00:01:13,150
>> Which brings us to the mission for
today's deep dive.
80
00:01:13,150 --> 00:01:13,160
today's deep dive.
81
00:01:13,160 --> 00:01:15,150
today's deep dive.
>> Mhm. We are looking at this incredible
82
00:01:15,150 --> 00:01:15,160
>> Mhm. We are looking at this incredible
83
00:01:15,160 --> 00:01:17,470
>> Mhm. We are looking at this incredible
2026 paper published in the journal
84
00:01:17,470 --> 00:01:17,480
2026 paper published in the journal
85
00:01:17,480 --> 00:01:19,670
2026 paper published in the journal
Cell. Yeah, a massive paper.
86
00:01:19,670 --> 00:01:19,680
Cell. Yeah, a massive paper.
87
00:01:19,680 --> 00:01:22,750
Cell. Yeah, a massive paper.
>> It really is. Uh, it's titled Reg-Velo,
88
00:01:22,750 --> 00:01:22,760
>> It really is. Uh, it's titled Reg-Velo,
89
00:01:22,760 --> 00:01:24,790
>> It really is. Uh, it's titled Reg-Velo,
Gene Regulatory Informed Dynamics of
90
00:01:24,790 --> 00:01:24,800
Gene Regulatory Informed Dynamics of
91
00:01:24,800 --> 00:01:27,430
Gene Regulatory Informed Dynamics of
Single Cells by Wang and a pretty large
92
00:01:27,430 --> 00:01:27,440
Single Cells by Wang and a pretty large
93
00:01:27,440 --> 00:01:28,350
Single Cells by Wang and a pretty large
team of researchers.
94
00:01:28,350 --> 00:01:28,360
team of researchers.
95
00:01:28,360 --> 00:01:30,550
team of researchers.
>> Yeah. And our goal today is to unpack
96
00:01:30,550 --> 00:01:30,560
>> Yeah. And our goal today is to unpack
97
00:01:30,560 --> 00:01:32,830
>> Yeah. And our goal today is to unpack
how this team has finally built like a
98
00:01:32,830 --> 00:01:32,840
how this team has finally built like a
99
00:01:32,840 --> 00:01:34,990
how this team has finally built like a
mathematical bridge between the speed at
100
00:01:34,990 --> 00:01:35,000
mathematical bridge between the speed at
101
00:01:35,000 --> 00:01:37,110
mathematical bridge between the speed at
which a cell changes and the internal
102
00:01:37,110 --> 00:01:37,120
which a cell changes and the internal
103
00:01:37,120 --> 00:01:38,710
which a cell changes and the internal
software that actually programs that
104
00:01:38,710 --> 00:01:38,720
software that actually programs that
105
00:01:38,720 --> 00:01:41,070
software that actually programs that
change. To fully grasp why Reg-Velo is
106
00:01:41,070 --> 00:01:41,080
change. To fully grasp why Reg-Velo is
107
00:01:41,080 --> 00:01:42,350
change. To fully grasp why Reg-Velo is
sending such shockwaves through
108
00:01:42,350 --> 00:01:42,360
sending such shockwaves through
109
00:01:42,360 --> 00:01:44,350
sending such shockwaves through
computational biology, we really have to
110
00:01:44,350 --> 00:01:44,360
computational biology, we really have to
111
00:01:44,360 --> 00:01:46,190
computational biology, we really have to
look at the massive blind spot that
112
00:01:46,190 --> 00:01:46,200
look at the massive blind spot that
113
00:01:46,200 --> 00:01:47,510
look at the massive blind spot that
researchers were dealing with before
114
00:01:47,510 --> 00:01:47,520
researchers were dealing with before
115
00:01:47,520 --> 00:01:48,990
researchers were dealing with before
this paper even came out.
116
00:01:48,990 --> 00:01:49,000
this paper even came out.
117
00:01:49,000 --> 00:01:50,350
this paper even came out.
>> Right. Historically, we had two
118
00:01:50,350 --> 00:01:50,360
>> Right. Historically, we had two
119
00:01:50,360 --> 00:01:53,390
>> Right. Historically, we had two
incredibly powerful, but um, completely
120
00:01:53,390 --> 00:01:53,400
incredibly powerful, but um, completely
121
00:01:53,400 --> 00:01:55,710
incredibly powerful, but um, completely
isolated tools for studying cells. Let's
122
00:01:55,710 --> 00:01:55,720
isolated tools for studying cells. Let's
123
00:01:55,720 --> 00:01:56,990
isolated tools for studying cells. Let's
lay those out for everyone listening.
124
00:01:56,990 --> 00:01:57,000
lay those out for everyone listening.
125
00:01:57,000 --> 00:01:58,830
lay those out for everyone listening.
The first one is a concept called RNA
126
00:01:58,830 --> 00:01:58,840
The first one is a concept called RNA
127
00:01:58,840 --> 00:01:59,590
The first one is a concept called RNA
velocity.
128
00:01:59,590 --> 00:01:59,600
velocity.
129
00:01:59,600 --> 00:02:01,910
velocity.
>> Yes. And I think of RNA velocity as a
130
00:02:01,910 --> 00:02:01,920
>> Yes. And I think of RNA velocity as a
131
00:02:01,920 --> 00:02:03,870
>> Yes. And I think of RNA velocity as a
well, like a cellular speedometer.
132
00:02:03,870 --> 00:02:03,880
well, like a cellular speedometer.
133
00:02:03,880 --> 00:02:05,030
well, like a cellular speedometer.
That's a great way to put it.
134
00:02:05,030 --> 00:02:05,040
That's a great way to put it.
135
00:02:05,040 --> 00:02:06,270
That's a great way to put it.
>> It doesn't just tell you what a cell is
136
00:02:06,270 --> 00:02:06,280
>> It doesn't just tell you what a cell is
137
00:02:06,280 --> 00:02:08,469
>> It doesn't just tell you what a cell is
doing right now. It measures the speed
138
00:02:08,469 --> 00:02:08,479
doing right now. It measures the speed
139
00:02:08,479 --> 00:02:09,590
doing right now. It measures the speed
and the direction of the cell's
140
00:02:09,590 --> 00:02:09,600
and the direction of the cell's
141
00:02:09,600 --> 00:02:11,630
and the direction of the cell's
development uh to predict what it's
142
00:02:11,630 --> 00:02:11,640
development uh to predict what it's
143
00:02:11,640 --> 00:02:12,510
development uh to predict what it's
going to do next.
144
00:02:12,510 --> 00:02:12,520
going to do next.
145
00:02:12,520 --> 00:02:14,790
going to do next.
>> Right. And the second tool is what we
146
00:02:14,790 --> 00:02:14,800
>> Right. And the second tool is what we
147
00:02:14,800 --> 00:02:18,070
>> Right. And the second tool is what we
call a gene regulatory network or a GRN.
148
00:02:18,070 --> 00:02:18,080
call a gene regulatory network or a GRN.
149
00:02:18,080 --> 00:02:20,630
call a gene regulatory network or a GRN.
Okay. Think of the GRN as the circuitry.
150
00:02:20,630 --> 00:02:20,640
Okay. Think of the GRN as the circuitry.
151
00:02:20,640 --> 00:02:22,949
Okay. Think of the GRN as the circuitry.
It is the genetic software map that
152
00:02:22,949 --> 00:02:22,959
It is the genetic software map that
153
00:02:22,959 --> 00:02:25,350
It is the genetic software map that
dictates, you know, which genes are
154
00:02:25,350 --> 00:02:25,360
dictates, you know, which genes are
155
00:02:25,360 --> 00:02:27,110
dictates, you know, which genes are
turning on and which genes are turning
156
00:02:27,110 --> 00:02:27,120
turning on and which genes are turning
157
00:02:27,120 --> 00:02:29,230
turning on and which genes are turning
off to actually drive the cell's
158
00:02:29,230 --> 00:02:29,240
off to actually drive the cell's
159
00:02:29,240 --> 00:02:30,110
off to actually drive the cell's
development.
160
00:02:30,110 --> 00:02:30,120
development.
161
00:02:30,120 --> 00:02:31,710
development.
>> So, we have a speedometer and we have a
162
00:02:31,710 --> 00:02:31,720
>> So, we have a speedometer and we have a
163
00:02:31,720 --> 00:02:32,710
>> So, we have a speedometer and we have a
wiring diagram.
164
00:02:32,710 --> 00:02:32,720
wiring diagram.
165
00:02:32,720 --> 00:02:33,550
wiring diagram.
>> Exactly.
166
00:02:33,550 --> 00:02:33,560
>> Exactly.
167
00:02:33,560 --> 00:02:34,830
>> Exactly.
>> But here is the crazy part.
168
00:02:34,830 --> 00:02:34,840
>> But here is the crazy part.
169
00:02:34,840 --> 00:02:37,430
>> But here is the crazy part.
>> Yeah. Before Reg-Velo, biologists had
170
00:02:37,430 --> 00:02:37,440
>> Yeah. Before Reg-Velo, biologists had
171
00:02:37,440 --> 00:02:39,190
>> Yeah. Before Reg-Velo, biologists had
absolutely no way to connect the two.
172
00:02:39,190 --> 00:02:39,200
absolutely no way to connect the two.
173
00:02:39,200 --> 00:02:41,750
absolutely no way to connect the two.
None. It's like imagine sitting in a car
174
00:02:41,750 --> 00:02:41,760
None. It's like imagine sitting in a car
175
00:02:41,760 --> 00:02:43,190
None. It's like imagine sitting in a car
with a speedometer that tells you you're
176
00:02:43,190 --> 00:02:43,200
with a speedometer that tells you you're
177
00:02:43,200 --> 00:02:44,590
with a speedometer that tells you you're
accelerating
178
00:02:44,590 --> 00:02:44,600
accelerating
179
00:02:44,600 --> 00:02:46,510
accelerating
and you're holding a schematic of the
180
00:02:46,510 --> 00:02:46,520
and you're holding a schematic of the
181
00:02:46,520 --> 00:02:48,070
and you're holding a schematic of the
engine's electrical wiring in your
182
00:02:48,070 --> 00:02:48,080
engine's electrical wiring in your
183
00:02:48,080 --> 00:02:51,110
engine's electrical wiring in your
hands, but you have no mechanical way to
184
00:02:51,110 --> 00:02:51,120
hands, but you have no mechanical way to
185
00:02:51,120 --> 00:02:53,110
hands, but you have no mechanical way to
see how pressing the gas pedal actually
186
00:02:53,110 --> 00:02:53,120
see how pressing the gas pedal actually
187
00:02:53,120 --> 00:02:54,230
see how pressing the gas pedal actually
changes the speed.
188
00:02:54,230 --> 00:02:54,240
changes the speed.
189
00:02:54,240 --> 00:02:56,150
changes the speed.
>> Yeah, it was a profound conceptual
190
00:02:56,150 --> 00:02:56,160
>> Yeah, it was a profound conceptual
191
00:02:56,160 --> 00:02:58,030
>> Yeah, it was a profound conceptual
disconnect. I mean, we could watch the
192
00:02:58,030 --> 00:02:58,040
disconnect. I mean, we could watch the
193
00:02:58,040 --> 00:02:59,949
disconnect. I mean, we could watch the
cells move along their developmental
194
00:02:59,949 --> 00:02:59,959
cells move along their developmental
195
00:02:59,959 --> 00:03:02,590
cells move along their developmental
paths and we could draw these static
196
00:03:02,590 --> 00:03:02,600
paths and we could draw these static
197
00:03:02,600 --> 00:03:05,030
paths and we could draw these static
maps of gene networks, but we just
198
00:03:05,030 --> 00:03:05,040
maps of gene networks, but we just
199
00:03:05,040 --> 00:03:06,590
maps of gene networks, but we just
couldn't link them mechanically. We had
200
00:03:06,590 --> 00:03:06,600
couldn't link them mechanically. We had
201
00:03:06,600 --> 00:03:09,509
couldn't link them mechanically. We had
literally no mathematical way to prove
202
00:03:09,509 --> 00:03:09,519
literally no mathematical way to prove
203
00:03:09,519 --> 00:03:11,670
literally no mathematical way to prove
how the wiring was driving the motion.
204
00:03:11,670 --> 00:03:11,680
how the wiring was driving the motion.
205
00:03:11,680 --> 00:03:13,270
how the wiring was driving the motion.
So, to understand how they finally
206
00:03:13,270 --> 00:03:13,280
So, to understand how they finally
207
00:03:13,280 --> 00:03:15,110
So, to understand how they finally
solved this, I feel like we really need
208
00:03:15,110 --> 00:03:15,120
solved this, I feel like we really need
209
00:03:15,120 --> 00:03:16,630
solved this, I feel like we really need
to lift the hood and look at how that
210
00:03:16,630 --> 00:03:16,640
to lift the hood and look at how that
211
00:03:16,640 --> 00:03:18,870
to lift the hood and look at how that
speedometer, the RNA velocity, actually
212
00:03:18,870 --> 00:03:18,880
speedometer, the RNA velocity, actually
213
00:03:18,880 --> 00:03:21,470
speedometer, the RNA velocity, actually
works. Good idea. Because from what the
214
00:03:21,470 --> 00:03:21,480
works. Good idea. Because from what the
215
00:03:21,480 --> 00:03:24,350
works. Good idea. Because from what the
sources show, the old way we calculated
216
00:03:24,350 --> 00:03:24,360
sources show, the old way we calculated
217
00:03:24,360 --> 00:03:26,509
sources show, the old way we calculated
it was frankly a little broken. Oh, it
218
00:03:26,509 --> 00:03:26,519
it was frankly a little broken. Oh, it
219
00:03:26,519 --> 00:03:28,910
it was frankly a little broken. Oh, it
was heavily flawed. But, you know, the
220
00:03:28,910 --> 00:03:28,920
was heavily flawed. But, you know, the
221
00:03:28,920 --> 00:03:31,390
was heavily flawed. But, you know, the
underlying biological mechanism of RNA
222
00:03:31,390 --> 00:03:31,400
underlying biological mechanism of RNA
223
00:03:31,400 --> 00:03:33,870
underlying biological mechanism of RNA
velocity is actually quite elegant.
224
00:03:33,870 --> 00:03:33,880
velocity is actually quite elegant.
225
00:03:33,880 --> 00:03:34,990
velocity is actually quite elegant.
Okay, break that down for us.
226
00:03:34,990 --> 00:03:35,000
Okay, break that down for us.
227
00:03:35,000 --> 00:03:37,430
Okay, break that down for us.
>> It relies on the life cycle of RNA
228
00:03:37,430 --> 00:03:37,440
>> It relies on the life cycle of RNA
229
00:03:37,440 --> 00:03:39,350
>> It relies on the life cycle of RNA
inside the cell. When a gene is
230
00:03:39,350 --> 00:03:39,360
inside the cell. When a gene is
231
00:03:39,360 --> 00:03:41,310
inside the cell. When a gene is
activated, it doesn't instantly produce
232
00:03:41,310 --> 00:03:41,320
activated, it doesn't instantly produce
233
00:03:41,320 --> 00:03:43,310
activated, it doesn't instantly produce
a perfect finished protein. Right.
234
00:03:43,310 --> 00:03:43,320
a perfect finished protein. Right.
235
00:03:43,320 --> 00:03:45,229
a perfect finished protein. Right.
>> First, it transcribes this rough draft,
236
00:03:45,229 --> 00:03:45,239
>> First, it transcribes this rough draft,
237
00:03:45,239 --> 00:03:47,670
>> First, it transcribes this rough draft,
which we call unspliced RNA. It's the
238
00:03:47,670 --> 00:03:47,680
which we call unspliced RNA. It's the
239
00:03:47,680 --> 00:03:49,710
which we call unspliced RNA. It's the
raw, newly minted instruction manual.
240
00:03:49,710 --> 00:03:49,720
raw, newly minted instruction manual.
241
00:03:49,720 --> 00:03:50,590
raw, newly minted instruction manual.
>> Exactly.
242
00:03:50,590 --> 00:03:50,600
>> Exactly.
243
00:03:50,600 --> 00:03:53,390
>> Exactly.
>> has all the um the biological junk in it
244
00:03:53,390 --> 00:03:53,400
>> has all the um the biological junk in it
245
00:03:53,400 --> 00:03:54,910
>> has all the um the biological junk in it
that hasn't been edited out yet.
246
00:03:54,910 --> 00:03:54,920
that hasn't been edited out yet.
247
00:03:54,920 --> 00:03:56,790
that hasn't been edited out yet.
>> Yes. And then the sales machinery comes
248
00:03:56,790 --> 00:03:56,800
>> Yes. And then the sales machinery comes
249
00:03:56,800 --> 00:03:59,350
>> Yes. And then the sales machinery comes
in, cuts out the unneeded parts, and
250
00:03:59,350 --> 00:03:59,360
in, cuts out the unneeded parts, and
251
00:03:59,360 --> 00:04:01,470
in, cuts out the unneeded parts, and
processes it into what we call spliced
252
00:04:01,470 --> 00:04:01,480
processes it into what we call spliced
253
00:04:01,480 --> 00:04:02,030
processes it into what we call spliced
RNA.
254
00:04:02,030 --> 00:04:02,040
RNA.
255
00:04:02,040 --> 00:04:02,590
RNA.
>> Okay, that was okay.
256
00:04:02,590 --> 00:04:02,600
>> Okay, that was okay.
257
00:04:02,600 --> 00:04:04,470
>> Okay, that was okay.
>> That is the mature, ready-to-use
258
00:04:04,470 --> 00:04:04,480
>> That is the mature, ready-to-use
259
00:04:04,480 --> 00:04:06,790
>> That is the mature, ready-to-use
instruction. This makes perfect sense
260
00:04:06,790 --> 00:04:06,800
instruction. This makes perfect sense
261
00:04:06,800 --> 00:04:08,510
instruction. This makes perfect sense
when you look at how scientists actually
262
00:04:08,510 --> 00:04:08,520
when you look at how scientists actually
263
00:04:08,520 --> 00:04:10,470
when you look at how scientists actually
use it to predict the future.
264
00:04:10,470 --> 00:04:10,480
use it to predict the future.
265
00:04:10,480 --> 00:04:13,750
use it to predict the future.
Like, by looking at the ratio of raw,
266
00:04:13,750 --> 00:04:13,760
Like, by looking at the ratio of raw,
267
00:04:13,760 --> 00:04:16,990
Like, by looking at the ratio of raw,
unspliced RNA to mature, spliced RNA,
268
00:04:16,990 --> 00:04:17,000
unspliced RNA to mature, spliced RNA,
269
00:04:17,000 --> 00:04:19,190
unspliced RNA to mature, spliced RNA,
you can tell if a gene is ramping up or
270
00:04:19,190 --> 00:04:19,200
you can tell if a gene is ramping up or
271
00:04:19,200 --> 00:04:20,110
you can tell if a gene is ramping up or
powering down.
272
00:04:20,110 --> 00:04:20,120
powering down.
273
00:04:20,120 --> 00:04:21,910
powering down.
>> Precisely. It reminds me of
274
00:04:21,910 --> 00:04:21,920
>> Precisely. It reminds me of
275
00:04:21,920 --> 00:04:23,710
>> Precisely. It reminds me of
walking into a kitchen. Ooh, I like
276
00:04:23,710 --> 00:04:23,720
walking into a kitchen. Ooh, I like
277
00:04:23,720 --> 00:04:24,670
walking into a kitchen. Ooh, I like
this.
278
00:04:24,670 --> 00:04:24,680
this.
279
00:04:24,680 --> 00:04:26,070
this.
If you look on the counter and see raw
280
00:04:26,070 --> 00:04:26,080
If you look on the counter and see raw
281
00:04:26,080 --> 00:04:27,750
If you look on the counter and see raw
ingredients, you know, flour, sugar,
282
00:04:27,750 --> 00:04:27,760
ingredients, you know, flour, sugar,
283
00:04:27,760 --> 00:04:29,350
ingredients, you know, flour, sugar,
carton of eggs,
284
00:04:29,350 --> 00:04:29,360
carton of eggs,
285
00:04:29,360 --> 00:04:30,630
carton of eggs,
and nothing is in the oven yet.
286
00:04:30,630 --> 00:04:30,640
and nothing is in the oven yet.
287
00:04:30,640 --> 00:04:32,630
and nothing is in the oven yet.
>> Right, you know what's coming. Exactly.
288
00:04:32,630 --> 00:04:32,640
>> Right, you know what's coming. Exactly.
289
00:04:32,640 --> 00:04:34,430
>> Right, you know what's coming. Exactly.
You can predict with near certainty that
290
00:04:34,430 --> 00:04:34,440
You can predict with near certainty that
291
00:04:34,440 --> 00:04:36,790
You can predict with near certainty that
the chef is about to bake something. You
292
00:04:36,790 --> 00:04:36,800
the chef is about to bake something. You
293
00:04:36,800 --> 00:04:38,270
the chef is about to bake something. You
are literally looking at the future
294
00:04:38,270 --> 00:04:38,280
are literally looking at the future
295
00:04:38,280 --> 00:04:40,190
are literally looking at the future
state of the kitchen by just measuring
296
00:04:40,190 --> 00:04:40,200
state of the kitchen by just measuring
297
00:04:40,200 --> 00:04:42,950
state of the kitchen by just measuring
the raw materials. That ratio of raw to
298
00:04:42,950 --> 00:04:42,960
the raw materials. That ratio of raw to
299
00:04:42,960 --> 00:04:45,350
the raw materials. That ratio of raw to
mature RNA allowed previous mathematical
300
00:04:45,350 --> 00:04:45,360
mature RNA allowed previous mathematical
301
00:04:45,360 --> 00:04:47,750
mature RNA allowed previous mathematical
models like Svelo to create these really
302
00:04:47,750 --> 00:04:47,760
models like Svelo to create these really
303
00:04:47,760 --> 00:04:49,870
models like Svelo to create these really
beautiful vector fields. They could
304
00:04:49,870 --> 00:04:49,880
beautiful vector fields. They could
305
00:04:49,880 --> 00:04:51,670
beautiful vector fields. They could
literally draw arrows showing exactly
306
00:04:51,670 --> 00:04:51,680
literally draw arrows showing exactly
307
00:04:51,680 --> 00:04:52,630
literally draw arrows showing exactly
where cell was heading.
308
00:04:52,630 --> 00:04:52,640
where cell was heading.
309
00:04:52,640 --> 00:04:54,909
where cell was heading.
>> Okay, but But, yeah, to make that math
310
00:04:54,909 --> 00:04:54,919
>> Okay, but But, yeah, to make that math
311
00:04:54,919 --> 00:04:56,909
>> Okay, but But, yeah, to make that math
work, those older models had to rely on
312
00:04:56,909 --> 00:04:56,919
work, those older models had to rely on
313
00:04:56,919 --> 00:04:59,190
work, those older models had to rely on
some incredibly restrictive and frankly
314
00:04:59,190 --> 00:04:59,200
some incredibly restrictive and frankly
315
00:04:59,200 --> 00:05:00,630
some incredibly restrictive and frankly
unrealistic assumptions.
316
00:05:00,630 --> 00:05:00,640
unrealistic assumptions.
317
00:05:00,640 --> 00:05:01,990
unrealistic assumptions.
>> This is where the old models hit a
318
00:05:01,990 --> 00:05:02,000
>> This is where the old models hit a
319
00:05:02,000 --> 00:05:03,790
>> This is where the old models hit a
massive wall.
320
00:05:03,790 --> 00:05:03,800
massive wall.
321
00:05:03,800 --> 00:05:05,430
massive wall.
What exactly were they assuming? Cuz I
322
00:05:05,430 --> 00:05:05,440
What exactly were they assuming? Cuz I
323
00:05:05,440 --> 00:05:07,270
What exactly were they assuming? Cuz I
know the math gets a little dense here.
324
00:05:07,270 --> 00:05:07,280
know the math gets a little dense here.
325
00:05:07,280 --> 00:05:09,390
know the math gets a little dense here.
Well, they assumed that every single
326
00:05:09,390 --> 00:05:09,400
Well, they assumed that every single
327
00:05:09,400 --> 00:05:11,510
Well, they assumed that every single
gene in the cell operated completely
328
00:05:11,510 --> 00:05:11,520
gene in the cell operated completely
329
00:05:11,520 --> 00:05:13,350
gene in the cell operated completely
independently of every other gene.
330
00:05:13,350 --> 00:05:13,360
independently of every other gene.
331
00:05:13,360 --> 00:05:14,270
independently of every other gene.
Independently.
332
00:05:14,270 --> 00:05:14,280
Independently.
333
00:05:14,280 --> 00:05:16,550
Independently.
>> Yeah. In mathematical terms, they used
334
00:05:16,550 --> 00:05:16,560
>> Yeah. In mathematical terms, they used
335
00:05:16,560 --> 00:05:18,830
>> Yeah. In mathematical terms, they used
what's called decoupled one-dimensional
336
00:05:18,830 --> 00:05:18,840
what's called decoupled one-dimensional
337
00:05:18,840 --> 00:05:22,230
what's called decoupled one-dimensional
ordinary differential equations, or ODEs
338
00:05:22,230 --> 00:05:22,240
ordinary differential equations, or ODEs
339
00:05:22,240 --> 00:05:24,190
ordinary differential equations, or ODEs
for short. Wait, wait, I need to stop
340
00:05:24,190 --> 00:05:24,200
for short. Wait, wait, I need to stop
341
00:05:24,200 --> 00:05:26,909
for short. Wait, wait, I need to stop
you there. They assumed genes operate in
342
00:05:26,909 --> 00:05:26,919
you there. They assumed genes operate in
343
00:05:26,919 --> 00:05:29,230
you there. They assumed genes operate in
a vacuum. Basically, yes. That's like
344
00:05:29,230 --> 00:05:29,240
a vacuum. Basically, yes. That's like
345
00:05:29,240 --> 00:05:30,830
a vacuum. Basically, yes. That's like
trying to understand city traffic by
346
00:05:30,830 --> 00:05:30,840
trying to understand city traffic by
347
00:05:30,840 --> 00:05:33,510
trying to understand city traffic by
focusing a camera on one single car,
348
00:05:33,510 --> 00:05:33,520
focusing a camera on one single car,
349
00:05:33,520 --> 00:05:35,590
focusing a camera on one single car,
while completely ignoring the traffic
350
00:05:35,590 --> 00:05:35,600
while completely ignoring the traffic
351
00:05:35,600 --> 00:05:37,550
while completely ignoring the traffic
lights, the stop signs, and the hundreds
352
00:05:37,550 --> 00:05:37,560
lights, the stop signs, and the hundreds
353
00:05:37,560 --> 00:05:39,310
lights, the stop signs, and the hundreds
of other drivers swerving around it.
354
00:05:39,310 --> 00:05:39,320
of other drivers swerving around it.
355
00:05:39,320 --> 00:05:42,190
of other drivers swerving around it.
>> It is a severe oversimplification, yes.
356
00:05:42,190 --> 00:05:42,200
>> It is a severe oversimplification, yes.
357
00:05:42,200 --> 00:05:44,590
>> It is a severe oversimplification, yes.
But, um, it was the only way they could
358
00:05:44,590 --> 00:05:44,600
But, um, it was the only way they could
359
00:05:44,600 --> 00:05:46,350
But, um, it was the only way they could
get the equations to resolve at the
360
00:05:46,350 --> 00:05:46,360
get the equations to resolve at the
361
00:05:46,360 --> 00:05:48,870
get the equations to resolve at the
time. Decoupled just means the equation
362
00:05:48,870 --> 00:05:48,880
time. Decoupled just means the equation
363
00:05:48,880 --> 00:05:50,910
time. Decoupled just means the equation
for gene A simply doesn't care what gene
364
00:05:50,910 --> 00:05:50,920
for gene A simply doesn't care what gene
365
00:05:50,920 --> 00:05:52,270
for gene A simply doesn't care what gene
B is doing.
366
00:05:52,270 --> 00:05:52,280
B is doing.
367
00:05:52,280 --> 00:05:54,030
B is doing.
And furthermore, they assumed the speed
368
00:05:54,030 --> 00:05:54,040
And furthermore, they assumed the speed
369
00:05:54,040 --> 00:05:55,510
And furthermore, they assumed the speed
at which that raw RNA was being
370
00:05:55,510 --> 00:05:55,520
at which that raw RNA was being
371
00:05:55,520 --> 00:05:57,030
at which that raw RNA was being
produced, the transcription rate, was
372
00:05:57,030 --> 00:05:57,040
produced, the transcription rate, was
373
00:05:57,040 --> 00:05:59,190
produced, the transcription rate, was
just a constant unchanging background
374
00:05:59,190 --> 00:05:59,200
just a constant unchanging background
375
00:05:59,200 --> 00:06:01,550
just a constant unchanging background
hum. The biology is basically the exact
376
00:06:01,550 --> 00:06:01,560
hum. The biology is basically the exact
377
00:06:01,560 --> 00:06:03,390
hum. The biology is basically the exact
opposite of a vacuum. Exactly.
378
00:06:03,390 --> 00:06:03,400
opposite of a vacuum. Exactly.
379
00:06:03,400 --> 00:06:05,190
opposite of a vacuum. Exactly.
>> It's a highly interactive, super messy
380
00:06:05,190 --> 00:06:05,200
>> It's a highly interactive, super messy
381
00:06:05,200 --> 00:06:07,510
>> It's a highly interactive, super messy
system. You have upstream genes acting
382
00:06:07,510 --> 00:06:07,520
system. You have upstream genes acting
383
00:06:07,520 --> 00:06:09,590
system. You have upstream genes acting
like, well, like traffic cops,
384
00:06:09,590 --> 00:06:09,600
like, well, like traffic cops,
385
00:06:09,600 --> 00:06:11,470
like, well, like traffic cops,
constantly turning downstream genes on
386
00:06:11,470 --> 00:06:11,480
constantly turning downstream genes on
387
00:06:11,480 --> 00:06:13,710
constantly turning downstream genes on
and off. And those upstream genes are
388
00:06:13,710 --> 00:06:13,720
and off. And those upstream genes are
389
00:06:13,720 --> 00:06:15,750
and off. And those upstream genes are
called transcription factors. By
390
00:06:15,750 --> 00:06:15,760
called transcription factors. By
391
00:06:15,760 --> 00:06:18,150
called transcription factors. By
excluding them, you know, by ignoring
392
00:06:18,150 --> 00:06:18,160
excluding them, you know, by ignoring
393
00:06:18,160 --> 00:06:19,710
excluding them, you know, by ignoring
the regulatory traffic cops that
394
00:06:19,710 --> 00:06:19,720
the regulatory traffic cops that
395
00:06:19,720 --> 00:06:22,150
the regulatory traffic cops that
actually drive a cell's differentiation,
396
00:06:22,150 --> 00:06:22,160
actually drive a cell's differentiation,
397
00:06:22,160 --> 00:06:24,670
actually drive a cell's differentiation,
those old RNA velocity models were
398
00:06:24,670 --> 00:06:24,680
those old RNA velocity models were
399
00:06:24,680 --> 00:06:27,190
those old RNA velocity models were
fundamentally missing the underlying
400
00:06:27,190 --> 00:06:27,200
fundamentally missing the underlying
401
00:06:27,200 --> 00:06:29,830
fundamentally missing the underlying
mechanism of Waddington's landscape.
402
00:06:29,830 --> 00:06:29,840
mechanism of Waddington's landscape.
403
00:06:29,840 --> 00:06:31,270
mechanism of Waddington's landscape.
They could see the marble rolling, but
404
00:06:31,270 --> 00:06:31,280
They could see the marble rolling, but
405
00:06:31,280 --> 00:06:33,390
They could see the marble rolling, but
they had absolutely no idea why it was
406
00:06:33,390 --> 00:06:33,400
they had absolutely no idea why it was
407
00:06:33,400 --> 00:06:33,830
they had absolutely no idea why it was
rolling.
408
00:06:33,830 --> 00:06:33,840
rolling.
409
00:06:33,840 --> 00:06:35,590
rolling.
>> Precisely. Okay. So, let's do a quick
410
00:06:35,590 --> 00:06:35,600
>> Precisely. Okay. So, let's do a quick
411
00:06:35,600 --> 00:06:36,870
>> Precisely. Okay. So, let's do a quick
recap for you listening, make sure we're
412
00:06:36,870 --> 00:06:36,880
recap for you listening, make sure we're
413
00:06:36,880 --> 00:06:37,710
recap for you listening, make sure we're
all on the same page.
414
00:06:37,710 --> 00:06:37,720
all on the same page.
415
00:06:37,720 --> 00:06:39,030
all on the same page.
>> Got you. Bullet one, we have
416
00:06:39,030 --> 00:06:39,040
>> Got you. Bullet one, we have
417
00:06:39,040 --> 00:06:40,350
>> Got you. Bullet one, we have
Waddington's landscape, which is our
418
00:06:40,350 --> 00:06:40,360
Waddington's landscape, which is our
419
00:06:40,360 --> 00:06:42,110
Waddington's landscape, which is our
mental image of how a cell decides its
420
00:06:42,110 --> 00:06:42,120
mental image of how a cell decides its
421
00:06:42,120 --> 00:06:44,390
mental image of how a cell decides its
fate, rolling down a hill. Yep. Bullet
422
00:06:44,390 --> 00:06:44,400
fate, rolling down a hill. Yep. Bullet
423
00:06:44,400 --> 00:06:46,070
fate, rolling down a hill. Yep. Bullet
two, we have RNA velocity, our
424
00:06:46,070 --> 00:06:46,080
two, we have RNA velocity, our
425
00:06:46,080 --> 00:06:48,390
two, we have RNA velocity, our
speedometer that predicts future cell
426
00:06:48,390 --> 00:06:48,400
speedometer that predicts future cell
427
00:06:48,400 --> 00:06:50,790
speedometer that predicts future cell
states by looking at raw versus mature
428
00:06:50,790 --> 00:06:50,800
states by looking at raw versus mature
429
00:06:50,800 --> 00:06:54,030
states by looking at raw versus mature
RNA in the kitchen. Perfect. And bullet
430
00:06:54,030 --> 00:06:54,040
RNA in the kitchen. Perfect. And bullet
431
00:06:54,040 --> 00:06:56,550
RNA in the kitchen. Perfect. And bullet
three, the old mathematical models
432
00:06:56,550 --> 00:06:56,560
three, the old mathematical models
433
00:06:56,560 --> 00:06:58,870
three, the old mathematical models
completely ignored the genetic wiring,
434
00:06:58,870 --> 00:06:58,880
completely ignored the genetic wiring,
435
00:06:58,880 --> 00:07:00,950
completely ignored the genetic wiring,
the GRNs, that connect all these
436
00:07:00,950 --> 00:07:00,960
the GRNs, that connect all these
437
00:07:00,960 --> 00:07:01,550
the GRNs, that connect all these
changes.
438
00:07:01,550 --> 00:07:01,560
changes.
439
00:07:01,560 --> 00:07:03,510
changes.
>> Because they use those decoupled
440
00:07:03,510 --> 00:07:03,520
>> Because they use those decoupled
441
00:07:03,520 --> 00:07:05,070
>> Because they use those decoupled
one-dimensional equations.
442
00:07:05,070 --> 00:07:05,080
one-dimensional equations.
443
00:07:05,080 --> 00:07:06,790
one-dimensional equations.
>> Right. So, the big question for you, the
444
00:07:06,790 --> 00:07:06,800
>> Right. So, the big question for you, the
445
00:07:06,800 --> 00:07:07,790
>> Right. So, the big question for you, the
listener,
446
00:07:07,790 --> 00:07:07,800
listener,
447
00:07:07,800 --> 00:07:10,510
listener,
is how do we actually fuse this wiring
448
00:07:10,510 --> 00:07:10,520
is how do we actually fuse this wiring
449
00:07:10,520 --> 00:07:12,590
is how do we actually fuse this wiring
map with the velocity prediction? How do
450
00:07:12,590 --> 00:07:12,600
map with the velocity prediction? How do
451
00:07:12,600 --> 00:07:14,110
map with the velocity prediction? How do
we build a better model? And that is
452
00:07:14,110 --> 00:07:14,120
we build a better model? And that is
453
00:07:14,120 --> 00:07:15,950
we build a better model? And that is
exactly where the authors of this paper
454
00:07:15,950 --> 00:07:15,960
exactly where the authors of this paper
455
00:07:15,960 --> 00:07:18,150
exactly where the authors of this paper
come in. They developed a framework
456
00:07:18,150 --> 00:07:18,160
come in. They developed a framework
457
00:07:18,160 --> 00:07:20,790
come in. They developed a framework
called RegVelo. RegVelo. It stands for
458
00:07:20,790 --> 00:07:20,800
called RegVelo. RegVelo. It stands for
459
00:07:20,800 --> 00:07:22,670
called RegVelo. RegVelo. It stands for
regulatory velocity.
460
00:07:22,670 --> 00:07:22,680
regulatory velocity.
461
00:07:22,680 --> 00:07:24,710
regulatory velocity.
And rather than relying on those old
462
00:07:24,710 --> 00:07:24,720
And rather than relying on those old
463
00:07:24,720 --> 00:07:26,150
And rather than relying on those old
decoupled equations we just talked
464
00:07:26,150 --> 00:07:26,160
decoupled equations we just talked
465
00:07:26,160 --> 00:07:28,590
decoupled equations we just talked
about, RegVelo is built on deep
466
00:07:28,590 --> 00:07:28,600
about, RegVelo is built on deep
467
00:07:28,600 --> 00:07:31,390
about, RegVelo is built on deep
generative modeling. Okay. It merges the
468
00:07:31,390 --> 00:07:31,400
generative modeling. Okay. It merges the
469
00:07:31,400 --> 00:07:33,510
generative modeling. Okay. It merges the
RNA velocity equations and the gene
470
00:07:33,510 --> 00:07:33,520
RNA velocity equations and the gene
471
00:07:33,520 --> 00:07:35,870
RNA velocity equations and the gene
regulatory networks into a single
472
00:07:35,870 --> 00:07:35,880
regulatory networks into a single
473
00:07:35,880 --> 00:07:37,830
regulatory networks into a single
high-dimensional ordinary differential
474
00:07:37,830 --> 00:07:37,840
high-dimensional ordinary differential
475
00:07:37,840 --> 00:07:38,510
high-dimensional ordinary differential
equation.
476
00:07:38,510 --> 00:07:38,520
equation.
477
00:07:38,520 --> 00:07:40,790
equation.
>> A high-dimensional equation. Okay, let's
478
00:07:40,790 --> 00:07:40,800
>> A high-dimensional equation. Okay, let's
479
00:07:40,800 --> 00:07:42,150
>> A high-dimensional equation. Okay, let's
translate that out of math speak for a
480
00:07:42,150 --> 00:07:42,160
translate that out of math speak for a
481
00:07:42,160 --> 00:07:42,550
translate that out of math speak for a
second.
482
00:07:42,550 --> 00:07:42,560
second.
483
00:07:42,560 --> 00:07:45,550
second.
>> Sure. If the old models were thousands
484
00:07:45,550 --> 00:07:45,560
>> Sure. If the old models were thousands
485
00:07:45,560 --> 00:07:48,550
>> Sure. If the old models were thousands
of isolated single-lane dirt roads, Reg
486
00:07:48,550 --> 00:07:48,560
of isolated single-lane dirt roads, Reg
487
00:07:48,560 --> 00:07:50,750
of isolated single-lane dirt roads, Reg
Velo is a mathematical superhighway
488
00:07:50,750 --> 00:07:50,760
Velo is a mathematical superhighway
489
00:07:50,760 --> 00:07:52,390
Velo is a mathematical superhighway
system where every single road
490
00:07:52,390 --> 00:07:52,400
system where every single road
491
00:07:52,400 --> 00:07:54,750
system where every single road
intersects. Oh, I see. Transcription is
492
00:07:54,750 --> 00:07:54,760
intersects. Oh, I see. Transcription is
493
00:07:54,760 --> 00:07:57,110
intersects. Oh, I see. Transcription is
no longer modeled as a constant isolated
494
00:07:57,110 --> 00:07:57,120
no longer modeled as a constant isolated
495
00:07:57,120 --> 00:07:59,590
no longer modeled as a constant isolated
event. It is now treated as a dynamic
496
00:07:59,590 --> 00:07:59,600
event. It is now treated as a dynamic
497
00:07:59,600 --> 00:08:01,310
event. It is now treated as a dynamic
process that depends entirely on the
498
00:08:01,310 --> 00:08:01,320
process that depends entirely on the
499
00:08:01,320 --> 00:08:03,390
process that depends entirely on the
regulations of other genes and the
500
00:08:03,390 --> 00:08:03,400
regulations of other genes and the
501
00:08:03,400 --> 00:08:05,550
regulations of other genes and the
passage of time. I found the mechanics
502
00:08:05,550 --> 00:08:05,560
passage of time. I found the mechanics
503
00:08:05,560 --> 00:08:06,950
passage of time. I found the mechanics
of how they actually built this
504
00:08:06,950 --> 00:08:06,960
of how they actually built this
505
00:08:06,960 --> 00:08:10,030
of how they actually built this
superhighway fascinating because um
506
00:08:10,030 --> 00:08:10,040
superhighway fascinating because um
507
00:08:10,040 --> 00:08:11,790
superhighway fascinating because um
Reg Velo doesn't just guess how these
508
00:08:11,790 --> 00:08:11,800
Reg Velo doesn't just guess how these
509
00:08:11,800 --> 00:08:13,350
Reg Velo doesn't just guess how these
genes connect out of thin air, right?
510
00:08:13,350 --> 00:08:13,360
genes connect out of thin air, right?
511
00:08:13,360 --> 00:08:14,470
genes connect out of thin air, right?
>> No, it can't.
512
00:08:14,470 --> 00:08:14,480
>> No, it can't.
513
00:08:14,480 --> 00:08:15,390
>> No, it can't.
>> a starting point.
514
00:08:15,390 --> 00:08:15,400
>> a starting point.
515
00:08:15,400 --> 00:08:16,710
>> a starting point.
>> Yeah. So, the scientists feed the
516
00:08:16,710 --> 00:08:16,720
>> Yeah. So, the scientists feed the
517
00:08:16,720 --> 00:08:18,710
>> Yeah. So, the scientists feed the
algorithm a starter map of the gene
518
00:08:18,710 --> 00:08:18,720
algorithm a starter map of the gene
519
00:08:18,720 --> 00:08:20,710
algorithm a starter map of the gene
regulatory network. Right. And they
520
00:08:20,710 --> 00:08:20,720
regulatory network. Right. And they
521
00:08:20,720 --> 00:08:22,870
regulatory network. Right. And they
derive that starter map from other types
522
00:08:22,870 --> 00:08:22,880
derive that starter map from other types
523
00:08:22,880 --> 00:08:25,110
derive that starter map from other types
of single-cell data. Often, they use
524
00:08:25,110 --> 00:08:25,120
of single-cell data. Often, they use
525
00:08:25,120 --> 00:08:27,350
of single-cell data. Often, they use
something called multiomic data.
526
00:08:27,350 --> 00:08:27,360
something called multiomic data.
527
00:08:27,360 --> 00:08:27,990
something called multiomic data.
>> Multiomic?
528
00:08:27,990 --> 00:08:28,000
>> Multiomic?
529
00:08:28,000 --> 00:08:29,590
>> Multiomic?
>> Which involves looking at the physical
530
00:08:29,590 --> 00:08:29,600
>> Which involves looking at the physical
531
00:08:29,600 --> 00:08:32,630
>> Which involves looking at the physical
accessibility of the DNA itself to guess
532
00:08:32,630 --> 00:08:32,640
accessibility of the DNA itself to guess
533
00:08:32,640 --> 00:08:34,350
accessibility of the DNA itself to guess
which genes might be able to talk to
534
00:08:34,350 --> 00:08:34,360
which genes might be able to talk to
535
00:08:34,360 --> 00:08:35,510
which genes might be able to talk to
each other.
536
00:08:35,510 --> 00:08:35,520
each other.
537
00:08:35,520 --> 00:08:37,310
each other.
We refer to these starting maps as
538
00:08:37,310 --> 00:08:37,320
We refer to these starting maps as
539
00:08:37,320 --> 00:08:38,990
We refer to these starting maps as
epigenetic priors.
540
00:08:38,990 --> 00:08:39,000
epigenetic priors.
541
00:08:39,000 --> 00:08:40,510
epigenetic priors.
>> Epigenetic priors.
542
00:08:40,510 --> 00:08:40,520
>> Epigenetic priors.
543
00:08:40,520 --> 00:08:42,510
>> Epigenetic priors.
>> Yeah. I love that concept. It's like
544
00:08:42,510 --> 00:08:42,520
>> Yeah. I love that concept. It's like
545
00:08:42,520 --> 00:08:44,110
>> Yeah. I love that concept. It's like
looking at the physical shape of the
546
00:08:44,110 --> 00:08:44,120
looking at the physical shape of the
547
00:08:44,120 --> 00:08:46,670
looking at the physical shape of the
DNA, like seeing which chapters of the
548
00:08:46,670 --> 00:08:46,680
DNA, like seeing which chapters of the
549
00:08:46,680 --> 00:08:48,870
DNA, like seeing which chapters of the
instruction manual are physically open
550
00:08:48,870 --> 00:08:48,880
instruction manual are physically open
551
00:08:48,880 --> 00:08:50,710
instruction manual are physically open
and readable to guess the wiring
552
00:08:50,710 --> 00:08:50,720
and readable to guess the wiring
553
00:08:50,720 --> 00:08:52,310
and readable to guess the wiring
diagram.
554
00:08:52,310 --> 00:08:52,320
diagram.
555
00:08:52,320 --> 00:08:52,590
diagram.
So,
556
00:08:52,590 --> 00:08:52,600
So,
557
00:08:52,600 --> 00:08:54,270
So,
>> take that rough starter map and they
558
00:08:54,270 --> 00:08:54,280
>> take that rough starter map and they
559
00:08:54,280 --> 00:08:56,710
>> take that rough starter map and they
feed it into a deep neural network. And
560
00:08:56,710 --> 00:08:56,720
feed it into a deep neural network. And
561
00:08:56,720 --> 00:08:58,190
feed it into a deep neural network. And
that neural network has to calculate
562
00:08:58,190 --> 00:08:58,200
that neural network has to calculate
563
00:08:58,200 --> 00:09:00,190
that neural network has to calculate
something absolutely critical called
564
00:09:00,190 --> 00:09:00,200
something absolutely critical called
565
00:09:00,200 --> 00:09:01,390
something absolutely critical called
latent time.
566
00:09:01,390 --> 00:09:01,400
latent time.
567
00:09:01,400 --> 00:09:02,829
latent time.
>> I'm really hoping we'd get to latent
568
00:09:02,829 --> 00:09:02,839
>> I'm really hoping we'd get to latent
569
00:09:02,839 --> 00:09:03,150
>> I'm really hoping we'd get to latent
time.
570
00:09:03,150 --> 00:09:03,160
time.
571
00:09:03,160 --> 00:09:04,310
time.
>> Oh, it's so cool.
572
00:09:04,310 --> 00:09:04,320
>> Oh, it's so cool.
573
00:09:04,320 --> 00:09:05,870
>> Oh, it's so cool.
>> Because when you scoop up a sample of
574
00:09:05,870 --> 00:09:05,880
>> Because when you scoop up a sample of
575
00:09:05,880 --> 00:09:07,750
>> Because when you scoop up a sample of
cells to study them, you aren't
576
00:09:07,750 --> 00:09:07,760
cells to study them, you aren't
577
00:09:07,760 --> 00:09:09,470
cells to study them, you aren't
capturing them all at the exact same
578
00:09:09,470 --> 00:09:09,480
capturing them all at the exact same
579
00:09:09,480 --> 00:09:10,710
capturing them all at the exact same
moment in their life cycle.
580
00:09:10,710 --> 00:09:10,720
moment in their life cycle.
581
00:09:10,720 --> 00:09:11,990
moment in their life cycle.
>> No, it's a mixed bag.
582
00:09:11,990 --> 00:09:12,000
>> No, it's a mixed bag.
583
00:09:12,000 --> 00:09:14,350
>> No, it's a mixed bag.
>> Right. So, latent time is like the
584
00:09:14,350 --> 00:09:14,360
>> Right. So, latent time is like the
585
00:09:14,360 --> 00:09:16,630
>> Right. So, latent time is like the
algorithm calculating an internal
586
00:09:16,630 --> 00:09:16,640
algorithm calculating an internal
587
00:09:16,640 --> 00:09:19,070
algorithm calculating an internal
biological clock for each individual
588
00:09:19,070 --> 00:09:19,080
biological clock for each individual
589
00:09:19,080 --> 00:09:21,150
biological clock for each individual
cell, placing them in perfect
590
00:09:21,150 --> 00:09:21,160
cell, placing them in perfect
591
00:09:21,160 --> 00:09:24,510
cell, placing them in perfect
chronological order along a timeline of
592
00:09:24,510 --> 00:09:24,520
chronological order along a timeline of
593
00:09:24,520 --> 00:09:27,190
chronological order along a timeline of
development. It is a crucial step. And
594
00:09:27,190 --> 00:09:27,200
development. It is a crucial step. And
595
00:09:27,200 --> 00:09:29,110
development. It is a crucial step. And
alongside that internal clock, the
596
00:09:29,110 --> 00:09:29,120
alongside that internal clock, the
597
00:09:29,120 --> 00:09:31,550
alongside that internal clock, the
neural network also generates a massive
598
00:09:31,550 --> 00:09:31,560
neural network also generates a massive
599
00:09:31,560 --> 00:09:34,470
neural network also generates a massive
matrix of weights to represent the
600
00:09:34,470 --> 00:09:34,480
matrix of weights to represent the
601
00:09:34,480 --> 00:09:36,030
matrix of weights to represent the
regulatory network.
602
00:09:36,030 --> 00:09:36,040
regulatory network.
603
00:09:36,040 --> 00:09:36,670
regulatory network.
>> of weights.
604
00:09:36,670 --> 00:09:36,680
>> of weights.
605
00:09:36,680 --> 00:09:38,830
>> of weights.
>> Yes. It is assigning a mathematical
606
00:09:38,830 --> 00:09:38,840
>> Yes. It is assigning a mathematical
607
00:09:38,840 --> 00:09:40,870
>> Yes. It is assigning a mathematical
score to how different genes interact
608
00:09:40,870 --> 00:09:40,880
score to how different genes interact
609
00:09:40,880 --> 00:09:42,630
score to how different genes interact
with each other. This weight matrix is
610
00:09:42,630 --> 00:09:42,640
with each other. This weight matrix is
611
00:09:42,640 --> 00:09:44,510
with each other. This weight matrix is
where this concept really clicked for me
612
00:09:44,510 --> 00:09:44,520
where this concept really clicked for me
613
00:09:44,520 --> 00:09:45,510
where this concept really clicked for me
when I was reading the paper.
614
00:09:45,510 --> 00:09:45,520
when I was reading the paper.
615
00:09:45,520 --> 00:09:47,590
when I was reading the paper.
>> Yeah. Yeah. If the neural network
616
00:09:47,590 --> 00:09:47,600
>> Yeah. Yeah. If the neural network
617
00:09:47,600 --> 00:09:49,830
>> Yeah. Yeah. If the neural network
observes the data and gives a connection
618
00:09:49,830 --> 00:09:49,840
observes the data and gives a connection
619
00:09:49,840 --> 00:09:52,790
observes the data and gives a connection
a positive number, it means gene A
620
00:09:52,790 --> 00:09:52,800
a positive number, it means gene A
621
00:09:52,800 --> 00:09:55,590
a positive number, it means gene A
actively turns on gene B. Yes. And if it
622
00:09:55,590 --> 00:09:55,600
actively turns on gene B. Yes. And if it
623
00:09:55,600 --> 00:09:57,390
actively turns on gene B. Yes. And if it
gives it a negative number, gene A
624
00:09:57,390 --> 00:09:57,400
gives it a negative number, gene A
625
00:09:57,400 --> 00:10:01,510
gives it a negative number, gene A
represses or shuts down gene B. Exactly.
626
00:10:01,510 --> 00:10:01,520
represses or shuts down gene B. Exactly.
627
00:10:01,520 --> 00:10:03,230
represses or shuts down gene B. Exactly.
And a zero means those two genes simply
628
00:10:03,230 --> 00:10:03,240
And a zero means those two genes simply
629
00:10:03,240 --> 00:10:05,310
And a zero means those two genes simply
don't interact at all.
630
00:10:05,310 --> 00:10:05,320
don't interact at all.
631
00:10:05,320 --> 00:10:08,070
don't interact at all.
Okay. But here is the true stroke of
632
00:10:08,070 --> 00:10:08,080
Okay. But here is the true stroke of
633
00:10:08,080 --> 00:10:10,390
Okay. But here is the true stroke of
genius in the Reg-Velo framework. What's
634
00:10:10,390 --> 00:10:10,400
genius in the Reg-Velo framework. What's
635
00:10:10,400 --> 00:10:12,550
genius in the Reg-Velo framework. What's
that? It knows that biological starter
636
00:10:12,550 --> 00:10:12,560
that? It knows that biological starter
637
00:10:12,560 --> 00:10:14,430
that? It knows that biological starter
maps, those epigenetic priors we talked
638
00:10:14,430 --> 00:10:14,440
maps, those epigenetic priors we talked
639
00:10:14,440 --> 00:10:16,350
maps, those epigenetic priors we talked
about, are notoriously noisy and
640
00:10:16,350 --> 00:10:16,360
about, are notoriously noisy and
641
00:10:16,360 --> 00:10:18,950
about, are notoriously noisy and
imperfect. It does not blindly trust the
642
00:10:18,950 --> 00:10:18,960
imperfect. It does not blindly trust the
643
00:10:18,960 --> 00:10:20,750
imperfect. It does not blindly trust the
prior information it was given. Because
644
00:10:20,750 --> 00:10:20,760
prior information it was given. Because
645
00:10:20,760 --> 00:10:22,510
prior information it was given. Because
in reality, a lot of those initial
646
00:10:22,510 --> 00:10:22,520
in reality, a lot of those initial
647
00:10:22,520 --> 00:10:24,550
in reality, a lot of those initial
guesses about which genes talk to each
648
00:10:24,550 --> 00:10:24,560
guesses about which genes talk to each
649
00:10:24,560 --> 00:10:25,910
guesses about which genes talk to each
other are just flat out wrong.
650
00:10:25,910 --> 00:10:25,920
other are just flat out wrong.
651
00:10:25,920 --> 00:10:27,750
other are just flat out wrong.
>> Exactly. Biological networks are
652
00:10:27,750 --> 00:10:27,760
>> Exactly. Biological networks are
653
00:10:27,760 --> 00:10:29,510
>> Exactly. Biological networks are
governed by a principle called sparsity.
654
00:10:29,510 --> 00:10:29,520
governed by a principle called sparsity.
655
00:10:29,520 --> 00:10:31,950
governed by a principle called sparsity.
>> Sparsity. Let's define that. It just
656
00:10:31,950 --> 00:10:31,960
>> Sparsity. Let's define that. It just
657
00:10:31,960 --> 00:10:33,830
>> Sparsity. Let's define that. It just
means not every gene interacts with
658
00:10:33,830 --> 00:10:33,840
means not every gene interacts with
659
00:10:33,840 --> 00:10:36,070
means not every gene interacts with
every other gene. Most only communicate
660
00:10:36,070 --> 00:10:36,080
every other gene. Most only communicate
661
00:10:36,080 --> 00:10:38,750
every other gene. Most only communicate
with a highly specific select few.
662
00:10:38,750 --> 00:10:38,760
with a highly specific select few.
663
00:10:38,760 --> 00:10:41,190
with a highly specific select few.
Right. So, Reg-Velo's deep learning
664
00:10:41,190 --> 00:10:41,200
Right. So, Reg-Velo's deep learning
665
00:10:41,200 --> 00:10:44,150
Right. So, Reg-Velo's deep learning
model uses the actual RNA velocity data
666
00:10:44,150 --> 00:10:44,160
model uses the actual RNA velocity data
667
00:10:44,160 --> 00:10:46,470
model uses the actual RNA velocity data
to learn new edge weights.
668
00:10:46,470 --> 00:10:46,480
to learn new edge weights.
669
00:10:46,480 --> 00:10:48,470
to learn new edge weights.
It essentially self-corrects.
670
00:10:48,470 --> 00:10:48,480
It essentially self-corrects.
671
00:10:48,480 --> 00:10:49,430
It essentially self-corrects.
>> That's wild.
672
00:10:49,430 --> 00:10:49,440
>> That's wild.
673
00:10:49,440 --> 00:10:51,870
>> That's wild.
>> Yeah. It fixes erroneous assumptions in
674
00:10:51,870 --> 00:10:51,880
>> Yeah. It fixes erroneous assumptions in
675
00:10:51,880 --> 00:10:53,710
>> Yeah. It fixes erroneous assumptions in
the original map and discovers hidden
676
00:10:53,710 --> 00:10:53,720
the original map and discovers hidden
677
00:10:53,720 --> 00:10:55,310
the original map and discovers hidden
connections that the researchers hadn't
678
00:10:55,310 --> 00:10:55,320
connections that the researchers hadn't
679
00:10:55,320 --> 00:10:56,270
connections that the researchers hadn't
even mapped yet.
680
00:10:56,270 --> 00:10:56,280
even mapped yet.
681
00:10:56,280 --> 00:10:59,070
even mapped yet.
>> It's rebuilding the car's wiring diagram
682
00:10:59,070 --> 00:10:59,080
>> It's rebuilding the car's wiring diagram
683
00:10:59,080 --> 00:11:00,910
>> It's rebuilding the car's wiring diagram
while the car is actively driving down
684
00:11:00,910 --> 00:11:00,920
while the car is actively driving down
685
00:11:00,920 --> 00:11:03,350
while the car is actively driving down
the highway. Yes. That's incredible. So,
686
00:11:03,350 --> 00:11:03,360
the highway. Yes. That's incredible. So,
687
00:11:03,360 --> 00:11:04,630
the highway. Yes. That's incredible. So,
time for another quick recap.
688
00:11:04,630 --> 00:11:04,640
time for another quick recap.
689
00:11:04,640 --> 00:11:07,190
time for another quick recap.
>> Let's do it. Bullet one, Reg-Velo merges
690
00:11:07,190 --> 00:11:07,200
>> Let's do it. Bullet one, Reg-Velo merges
691
00:11:07,200 --> 00:11:09,390
>> Let's do it. Bullet one, Reg-Velo merges
RNA velocity with those gene regulatory
692
00:11:09,390 --> 00:11:09,400
RNA velocity with those gene regulatory
693
00:11:09,400 --> 00:11:10,030
RNA velocity with those gene regulatory
networks.
694
00:11:10,030 --> 00:11:10,040
networks.
695
00:11:10,040 --> 00:11:12,270
networks.
>> Yeah. Bullet two, it uses complex
696
00:11:12,270 --> 00:11:12,280
>> Yeah. Bullet two, it uses complex
697
00:11:12,280 --> 00:11:14,150
>> Yeah. Bullet two, it uses complex
coupled math, those high-dimensional
698
00:11:14,150 --> 00:11:14,160
coupled math, those high-dimensional
699
00:11:14,160 --> 00:11:16,430
coupled math, those high-dimensional
ODEs, instead of isolated equations.
700
00:11:16,430 --> 00:11:16,440
ODEs, instead of isolated equations.
701
00:11:16,440 --> 00:11:18,430
ODEs, instead of isolated equations.
>> Right. And bullet three, it refines
702
00:11:18,430 --> 00:11:18,440
>> Right. And bullet three, it refines
703
00:11:18,440 --> 00:11:20,790
>> Right. And bullet three, it refines
imperfect prior knowledge using deep
704
00:11:20,790 --> 00:11:20,800
imperfect prior knowledge using deep
705
00:11:20,800 --> 00:11:22,829
imperfect prior knowledge using deep
learning, keeping sparsity in mind.
706
00:11:22,829 --> 00:11:22,839
learning, keeping sparsity in mind.
707
00:11:22,839 --> 00:11:24,870
learning, keeping sparsity in mind.
Perfectly summarized. So, for you
708
00:11:24,870 --> 00:11:24,880
Perfectly summarized. So, for you
709
00:11:24,880 --> 00:11:26,590
Perfectly summarized. So, for you
listening, think about this question. If
710
00:11:26,590 --> 00:11:26,600
listening, think about this question. If
711
00:11:26,600 --> 00:11:28,430
listening, think about this question. If
we have this highly accurate,
712
00:11:28,430 --> 00:11:28,440
we have this highly accurate,
713
00:11:28,440 --> 00:11:30,870
we have this highly accurate,
mathematically coupled virtual model,
714
00:11:30,870 --> 00:11:30,880
mathematically coupled virtual model,
715
00:11:30,880 --> 00:11:32,510
mathematically coupled virtual model,
what happens if we start tinkering with
716
00:11:32,510 --> 00:11:32,520
what happens if we start tinkering with
717
00:11:32,520 --> 00:11:34,750
what happens if we start tinkering with
it? And this is where Reg Velo
718
00:11:34,750 --> 00:11:34,760
it? And this is where Reg Velo
719
00:11:34,760 --> 00:11:36,310
it? And this is where Reg Velo
transitions from being just an
720
00:11:36,310 --> 00:11:36,320
transitions from being just an
721
00:11:36,320 --> 00:11:38,710
transitions from being just an
observational tool to being a full-on
722
00:11:38,710 --> 00:11:38,720
observational tool to being a full-on
723
00:11:38,720 --> 00:11:40,150
observational tool to being a full-on
experimental playground.
724
00:11:40,150 --> 00:11:40,160
experimental playground.
725
00:11:40,160 --> 00:11:41,630
experimental playground.
>> Oh, this part is so cool. The
726
00:11:41,630 --> 00:11:41,640
>> Oh, this part is so cool. The
727
00:11:41,640 --> 00:11:43,470
>> Oh, this part is so cool. The
researchers connected Reg Velo to an
728
00:11:43,470 --> 00:11:43,480
researchers connected Reg Velo to an
729
00:11:43,480 --> 00:11:46,430
researchers connected Reg Velo to an
analytical tool called Cell Rank 2. Cell
730
00:11:46,430 --> 00:11:46,440
analytical tool called Cell Rank 2. Cell
731
00:11:46,440 --> 00:11:48,670
analytical tool called Cell Rank 2. Cell
Rank 2? And this combination allows
732
00:11:48,670 --> 00:11:48,680
Rank 2? And this combination allows
733
00:11:48,680 --> 00:11:50,790
Rank 2? And this combination allows
scientists to perform what are known as
734
00:11:50,790 --> 00:11:50,800
scientists to perform what are known as
735
00:11:50,800 --> 00:11:52,790
scientists to perform what are known as
in silico perturbations.
736
00:11:52,790 --> 00:11:52,800
in silico perturbations.
737
00:11:52,800 --> 00:11:54,790
in silico perturbations.
>> In silico meaning entirely inside the
738
00:11:54,790 --> 00:11:54,800
>> In silico meaning entirely inside the
739
00:11:54,800 --> 00:11:57,550
>> In silico meaning entirely inside the
computer. Exactly. They have essentially
740
00:11:57,550 --> 00:11:57,560
computer. Exactly. They have essentially
741
00:11:57,560 --> 00:12:00,990
computer. Exactly. They have essentially
constructed a fully actionable virtual
742
00:12:00,990 --> 00:12:01,000
constructed a fully actionable virtual
743
00:12:01,000 --> 00:12:02,110
constructed a fully actionable virtual
cell.
744
00:12:02,110 --> 00:12:02,120
cell.
745
00:12:02,120 --> 00:12:03,430
cell.
I'm trying to picture what this looks
746
00:12:03,430 --> 00:12:03,440
I'm trying to picture what this looks
747
00:12:03,440 --> 00:12:05,590
I'm trying to picture what this looks
like in practice because if I'm a
748
00:12:05,590 --> 00:12:05,600
like in practice because if I'm a
749
00:12:05,600 --> 00:12:07,630
like in practice because if I'm a
researcher studying a specific organ,
750
00:12:07,630 --> 00:12:07,640
researcher studying a specific organ,
751
00:12:07,640 --> 00:12:09,790
researcher studying a specific organ,
normally I'd spend what, a year breeding
752
00:12:09,790 --> 00:12:09,800
normally I'd spend what, a year breeding
753
00:12:09,800 --> 00:12:11,270
normally I'd spend what, a year breeding
mice? Oh, easily.
754
00:12:11,270 --> 00:12:11,280
mice? Oh, easily.
755
00:12:11,280 --> 00:12:13,350
mice? Oh, easily.
>> Chemically altering their DNA and
756
00:12:13,350 --> 00:12:13,360
>> Chemically altering their DNA and
757
00:12:13,360 --> 00:12:14,750
>> Chemically altering their DNA and
waiting to see what a specific gene
758
00:12:14,750 --> 00:12:14,760
waiting to see what a specific gene
759
00:12:14,760 --> 00:12:15,750
waiting to see what a specific gene
mutation does.
760
00:12:15,750 --> 00:12:15,760
mutation does.
761
00:12:15,760 --> 00:12:17,670
mutation does.
>> Yeah. But with this, am I literally just
762
00:12:17,670 --> 00:12:17,680
>> Yeah. But with this, am I literally just
763
00:12:17,680 --> 00:12:19,670
>> Yeah. But with this, am I literally just
typing code to delete a gene and
764
00:12:19,670 --> 00:12:19,680
typing code to delete a gene and
765
00:12:19,680 --> 00:12:21,510
typing code to delete a gene and
watching the virtual organ break?
766
00:12:21,510 --> 00:12:21,520
watching the virtual organ break?
767
00:12:21,520 --> 00:12:23,030
watching the virtual organ break?
>> That's exactly what you were doing. You
768
00:12:23,030 --> 00:12:23,040
>> That's exactly what you were doing. You
769
00:12:23,040 --> 00:12:24,990
>> That's exactly what you were doing. You
go into the matrix, you digitally erase
770
00:12:24,990 --> 00:12:25,000
go into the matrix, you digitally erase
771
00:12:25,000 --> 00:12:26,390
go into the matrix, you digitally erase
a gene and all of its downstream
772
00:12:26,390 --> 00:12:26,400
a gene and all of its downstream
773
00:12:26,400 --> 00:12:28,750
a gene and all of its downstream
targets, you run a forward simulation,
774
00:12:28,750 --> 00:12:28,760
targets, you run a forward simulation,
775
00:12:28,760 --> 00:12:30,190
targets, you run a forward simulation,
and you watch the mathematical
776
00:12:30,190 --> 00:12:30,200
and you watch the mathematical
777
00:12:30,200 --> 00:12:32,110
and you watch the mathematical
consequences ripple through the cell's
778
00:12:32,110 --> 00:12:32,120
consequences ripple through the cell's
779
00:12:32,120 --> 00:12:34,590
consequences ripple through the cell's
fate. Wow. It allows for counterfactual
780
00:12:34,590 --> 00:12:34,600
fate. Wow. It allows for counterfactual
781
00:12:34,600 --> 00:12:36,310
fate. Wow. It allows for counterfactual
inference. You know, asking the model,
782
00:12:36,310 --> 00:12:36,320
inference. You know, asking the model,
783
00:12:36,320 --> 00:12:37,870
inference. You know, asking the model,
"What would happen if this specific
784
00:12:37,870 --> 00:12:37,880
"What would happen if this specific
785
00:12:37,880 --> 00:12:40,190
"What would happen if this specific
traffic cop disappeared?" So, to prove
786
00:12:40,190 --> 00:12:40,200
traffic cop disappeared?" So, to prove
787
00:12:40,200 --> 00:12:42,470
traffic cop disappeared?" So, to prove
this wasn't just a fancy math trick, the
788
00:12:42,470 --> 00:12:42,480
this wasn't just a fancy math trick, the
789
00:12:42,480 --> 00:12:44,350
this wasn't just a fancy math trick, the
team obviously had to test it on some
790
00:12:44,350 --> 00:12:44,360
team obviously had to test it on some
791
00:12:44,360 --> 00:12:46,470
team obviously had to test it on some
highly researched biological systems.
792
00:12:46,470 --> 00:12:46,480
highly researched biological systems.
793
00:12:46,480 --> 00:12:48,350
highly researched biological systems.
Naturally. And I know they started with
794
00:12:48,350 --> 00:12:48,360
Naturally. And I know they started with
795
00:12:48,360 --> 00:12:49,910
Naturally. And I know they started with
mouse pancreatic development.
796
00:12:49,910 --> 00:12:49,920
mouse pancreatic development.
797
00:12:49,920 --> 00:12:51,550
mouse pancreatic development.
>> Yeah, the pancreas is a great testing
798
00:12:51,550 --> 00:12:51,560
>> Yeah, the pancreas is a great testing
799
00:12:51,560 --> 00:12:52,870
>> Yeah, the pancreas is a great testing
ground because we already know a lot
800
00:12:52,870 --> 00:12:52,880
ground because we already know a lot
801
00:12:52,880 --> 00:12:54,550
ground because we already know a lot
about how its different cells develop,
802
00:12:54,550 --> 00:12:54,560
about how its different cells develop,
803
00:12:54,560 --> 00:12:56,230
about how its different cells develop,
like the insulin-producing beta cells
804
00:12:56,230 --> 00:12:56,240
like the insulin-producing beta cells
805
00:12:56,240 --> 00:12:57,430
like the insulin-producing beta cells
and the ductal cells.
806
00:12:57,430 --> 00:12:57,440
and the ductal cells.
807
00:12:57,440 --> 00:12:59,350
and the ductal cells.
>> Right. When they fed the raw pancreatic
808
00:12:59,350 --> 00:12:59,360
>> Right. When they fed the raw pancreatic
809
00:12:59,360 --> 00:13:01,550
>> Right. When they fed the raw pancreatic
data into Reg Velo, the model
810
00:13:01,550 --> 00:13:01,560
data into Reg Velo, the model
811
00:13:01,560 --> 00:13:03,710
data into Reg Velo, the model
independently identified the known
812
00:13:03,710 --> 00:13:03,720
independently identified the known
813
00:13:03,720 --> 00:13:05,030
independently identified the known
driver genes.
814
00:13:05,030 --> 00:13:05,040
driver genes.
815
00:13:05,040 --> 00:13:07,430
driver genes.
>> Which is a huge first step. And then,
816
00:13:07,430 --> 00:13:07,440
>> Which is a huge first step. And then,
817
00:13:07,440 --> 00:13:09,110
>> Which is a huge first step. And then,
when they did a virtual knockout of a
818
00:13:09,110 --> 00:13:09,120
when they did a virtual knockout of a
819
00:13:09,120 --> 00:13:11,670
when they did a virtual knockout of a
gene called E2F1,
820
00:13:11,670 --> 00:13:11,680
gene called E2F1,
821
00:13:11,680 --> 00:13:14,430
gene called E2F1,
>> Yes. the simulation instantly showed a
822
00:13:14,430 --> 00:13:14,440
>> Yes. the simulation instantly showed a
823
00:13:14,440 --> 00:13:16,950
>> Yes. the simulation instantly showed a
massive disruption in the cell cycle of
824
00:13:16,950 --> 00:13:16,960
massive disruption in the cell cycle of
825
00:13:16,960 --> 00:13:17,870
massive disruption in the cell cycle of
the ductal cells.
826
00:13:17,870 --> 00:13:17,880
the ductal cells.
827
00:13:17,880 --> 00:13:19,230
the ductal cells.
>> Exactly. And then when they digitally
828
00:13:19,230 --> 00:13:19,240
>> Exactly. And then when they digitally
829
00:13:19,240 --> 00:13:21,750
>> Exactly. And then when they digitally
knocked out a gene called RFF6, the
830
00:13:21,750 --> 00:13:21,760
knocked out a gene called RFF6, the
831
00:13:21,760 --> 00:13:23,150
knocked out a gene called RFF6, the
virtual cell suddenly failed to
832
00:13:23,150 --> 00:13:23,160
virtual cell suddenly failed to
833
00:13:23,160 --> 00:13:25,110
virtual cell suddenly failed to
differentiate into beta cells. It
834
00:13:25,110 --> 00:13:25,120
differentiate into beta cells. It
835
00:13:25,120 --> 00:13:27,190
differentiate into beta cells. It
perfectly mirrored decades of hard
836
00:13:27,190 --> 00:13:27,200
perfectly mirrored decades of hard
837
00:13:27,200 --> 00:13:28,870
perfectly mirrored decades of hard
thought, real-world biological
838
00:13:28,870 --> 00:13:28,880
thought, real-world biological
839
00:13:28,880 --> 00:13:31,150
thought, real-world biological
experiments just by crunching the data.
840
00:13:31,150 --> 00:13:31,160
experiments just by crunching the data.
841
00:13:31,160 --> 00:13:32,830
experiments just by crunching the data.
But an even more striking example from
842
00:13:32,830 --> 00:13:32,840
But an even more striking example from
843
00:13:32,840 --> 00:13:35,390
But an even more striking example from
the paper involves human hematopoiesis.
844
00:13:35,390 --> 00:13:35,400
the paper involves human hematopoiesis.
845
00:13:35,400 --> 00:13:37,630
the paper involves human hematopoiesis.
The formation of blood cells, yes. This
846
00:13:37,630 --> 00:13:37,640
The formation of blood cells, yes. This
847
00:13:37,640 --> 00:13:39,870
The formation of blood cells, yes. This
system relies on a classic textbook
848
00:13:39,870 --> 00:13:39,880
system relies on a classic textbook
849
00:13:39,880 --> 00:13:41,910
system relies on a classic textbook
example of gene regulation known as a
850
00:13:41,910 --> 00:13:41,920
example of gene regulation known as a
851
00:13:41,920 --> 00:13:43,550
example of gene regulation known as a
toggle switch. This is one of my
852
00:13:43,550 --> 00:13:43,560
toggle switch. This is one of my
853
00:13:43,560 --> 00:13:44,590
toggle switch. This is one of my
favorite parts of the paper.
854
00:13:44,590 --> 00:13:44,600
favorite parts of the paper.
855
00:13:44,600 --> 00:13:46,110
favorite parts of the paper.
>> Let's break it down. It's the tug-of-war
856
00:13:46,110 --> 00:13:46,120
>> Let's break it down. It's the tug-of-war
857
00:13:46,120 --> 00:13:49,630
>> Let's break it down. It's the tug-of-war
between GATA1 and SPI1. Yes. And SPI1 is
858
00:13:49,630 --> 00:13:49,640
between GATA1 and SPI1. Yes. And SPI1 is
859
00:13:49,640 --> 00:13:51,670
between GATA1 and SPI1. Yes. And SPI1 is
also commonly known as PU.1 just for
860
00:13:51,670 --> 00:13:51,680
also commonly known as PU.1 just for
861
00:13:51,680 --> 00:13:52,990
also commonly known as PU.1 just for
anyone who might know it by that name.
862
00:13:52,990 --> 00:13:53,000
anyone who might know it by that name.
863
00:13:53,000 --> 00:13:56,710
anyone who might know it by that name.
>> Okay, so GATA1 versus SPI1. When a human
864
00:13:56,710 --> 00:13:56,720
>> Okay, so GATA1 versus SPI1. When a human
865
00:13:56,720 --> 00:13:59,030
>> Okay, so GATA1 versus SPI1. When a human
stem cell is preparing to become a blood
866
00:13:59,030 --> 00:13:59,040
stem cell is preparing to become a blood
867
00:13:59,040 --> 00:14:01,830
stem cell is preparing to become a blood
cell, it faces a critical fork in
868
00:14:01,830 --> 00:14:01,840
cell, it faces a critical fork in
869
00:14:01,840 --> 00:14:03,910
cell, it faces a critical fork in
Waddington's landscape. Right. It can
870
00:14:03,910 --> 00:14:03,920
Waddington's landscape. Right. It can
871
00:14:03,920 --> 00:14:05,830
Waddington's landscape. Right. It can
slide down one valley to become an
872
00:14:05,830 --> 00:14:05,840
slide down one valley to become an
873
00:14:05,840 --> 00:14:07,950
slide down one valley to become an
erythroid cell, which is a red blood
874
00:14:07,950 --> 00:14:07,960
erythroid cell, which is a red blood
875
00:14:07,960 --> 00:14:10,470
erythroid cell, which is a red blood
cell that carries oxygen. Okay. Or it
876
00:14:10,470 --> 00:14:10,480
cell that carries oxygen. Okay. Or it
877
00:14:10,480 --> 00:14:12,190
cell that carries oxygen. Okay. Or it
can slide down a completely different
878
00:14:12,190 --> 00:14:12,200
can slide down a completely different
879
00:14:12,200 --> 00:14:14,470
can slide down a completely different
valley to become a monocyte, which is a
880
00:14:14,470 --> 00:14:14,480
valley to become a monocyte, which is a
881
00:14:14,480 --> 00:14:15,990
valley to become a monocyte, which is a
white blood cell that fights off
882
00:14:15,990 --> 00:14:16,000
white blood cell that fights off
883
00:14:16,000 --> 00:14:17,990
white blood cell that fights off
infections. And the thing deciding which
884
00:14:17,990 --> 00:14:18,000
infections. And the thing deciding which
885
00:14:18,000 --> 00:14:21,310
infections. And the thing deciding which
valley the cell falls into is a literal
886
00:14:21,310 --> 00:14:21,320
valley the cell falls into is a literal
887
00:14:21,320 --> 00:14:23,150
valley the cell falls into is a literal
biological tug-of-war between those two
888
00:14:23,150 --> 00:14:23,160
biological tug-of-war between those two
889
00:14:23,160 --> 00:14:25,750
biological tug-of-war between those two
transcription factors. Exactly. GATA1
890
00:14:25,750 --> 00:14:25,760
transcription factors. Exactly. GATA1
891
00:14:25,760 --> 00:14:28,150
transcription factors. Exactly. GATA1
and SPI1 mutually repress each other. So
892
00:14:28,150 --> 00:14:28,160
and SPI1 mutually repress each other. So
893
00:14:28,160 --> 00:14:30,230
and SPI1 mutually repress each other. So
if GATA1 gets the upper hand, it
894
00:14:30,230 --> 00:14:30,240
if GATA1 gets the upper hand, it
895
00:14:30,240 --> 00:14:33,070
if GATA1 gets the upper hand, it
physically blocks SPI1 from functioning
896
00:14:33,070 --> 00:14:33,080
physically blocks SPI1 from functioning
897
00:14:33,080 --> 00:14:34,710
physically blocks SPI1 from functioning
and the cell becomes a red blood cell.
898
00:14:34,710 --> 00:14:34,720
and the cell becomes a red blood cell.
899
00:14:34,720 --> 00:14:38,110
and the cell becomes a red blood cell.
Yes. And if SPI1 wins the tug-of-war, it
900
00:14:38,110 --> 00:14:38,120
Yes. And if SPI1 wins the tug-of-war, it
901
00:14:38,120 --> 00:14:40,350
Yes. And if SPI1 wins the tug-of-war, it
shuts down GATA1 and the cell becomes a
902
00:14:40,350 --> 00:14:40,360
shuts down GATA1 and the cell becomes a
903
00:14:40,360 --> 00:14:42,350
shuts down GATA1 and the cell becomes a
monocyte. What is so impressive is that
904
00:14:42,350 --> 00:14:42,360
monocyte. What is so impressive is that
905
00:14:42,360 --> 00:14:44,230
monocyte. What is so impressive is that
Rigvel captured this toggle switch
906
00:14:44,230 --> 00:14:44,240
Rigvel captured this toggle switch
907
00:14:44,240 --> 00:14:46,430
Rigvel captured this toggle switch
perfectly. The researchers went into the
908
00:14:46,430 --> 00:14:46,440
perfectly. The researchers went into the
909
00:14:46,440 --> 00:14:48,670
perfectly. The researchers went into the
virtual model, computationally deleted
910
00:14:48,670 --> 00:14:48,680
virtual model, computationally deleted
911
00:14:48,680 --> 00:14:50,710
virtual model, computationally deleted
the GATA1 network, and they watched the
912
00:14:50,710 --> 00:14:50,720
the GATA1 network, and they watched the
913
00:14:50,720 --> 00:14:52,590
the GATA1 network, and they watched the
mathematical fate probabilities
914
00:14:52,590 --> 00:14:52,600
mathematical fate probabilities
915
00:14:52,600 --> 00:14:54,510
mathematical fate probabilities
instantly shift entirely toward the
916
00:14:54,510 --> 00:14:54,520
instantly shift entirely toward the
917
00:14:54,520 --> 00:14:57,750
instantly shift entirely toward the
monocyte lineage. It's a perfect digital
918
00:14:57,750 --> 00:14:57,760
monocyte lineage. It's a perfect digital
919
00:14:57,760 --> 00:15:00,670
monocyte lineage. It's a perfect digital
mirror of biological reality. It really
920
00:15:00,670 --> 00:15:00,680
mirror of biological reality. It really
921
00:15:00,680 --> 00:15:02,829
mirror of biological reality. It really
is. And it's worth noting that the older
922
00:15:02,829 --> 00:15:02,839
is. And it's worth noting that the older
923
00:15:02,839 --> 00:15:05,230
is. And it's worth noting that the older
decoupled RNA velocity models totally
924
00:15:05,230 --> 00:15:05,240
decoupled RNA velocity models totally
925
00:15:05,240 --> 00:15:07,470
decoupled RNA velocity models totally
failed at this. Miserably. Because they
926
00:15:07,470 --> 00:15:07,480
failed at this. Miserably. Because they
927
00:15:07,480 --> 00:15:09,310
failed at this. Miserably. Because they
couldn't see the regulatory traffic
928
00:15:09,310 --> 00:15:09,320
couldn't see the regulatory traffic
929
00:15:09,320 --> 00:15:10,470
couldn't see the regulatory traffic
cops, they couldn't predict the
930
00:15:10,470 --> 00:15:10,480
cops, they couldn't predict the
931
00:15:10,480 --> 00:15:12,870
cops, they couldn't predict the
tug-of-war. Even when previous models
932
00:15:12,870 --> 00:15:12,880
tug-of-war. Even when previous models
933
00:15:12,880 --> 00:15:15,550
tug-of-war. Even when previous models
were fed extra highly detailed metabolic
934
00:15:15,550 --> 00:15:15,560
were fed extra highly detailed metabolic
935
00:15:15,560 --> 00:15:17,550
were fed extra highly detailed metabolic
data, they still couldn't accurately
936
00:15:17,550 --> 00:15:17,560
data, they still couldn't accurately
937
00:15:17,560 --> 00:15:19,310
data, they still couldn't accurately
predict these driver genes because they
938
00:15:19,310 --> 00:15:19,320
predict these driver genes because they
939
00:15:19,320 --> 00:15:21,390
predict these driver genes because they
fundamentally lacked the regulatory
940
00:15:21,390 --> 00:15:21,400
fundamentally lacked the regulatory
941
00:15:21,400 --> 00:15:22,190
fundamentally lacked the regulatory
context.
942
00:15:22,190 --> 00:15:22,200
context.
943
00:15:22,200 --> 00:15:24,070
context.
>> That's the power of coupling the math.
944
00:15:24,070 --> 00:15:24,080
>> That's the power of coupling the math.
945
00:15:24,080 --> 00:15:25,990
>> That's the power of coupling the math.
Okay, time for recap number three. Go
946
00:15:25,990 --> 00:15:26,000
Okay, time for recap number three. Go
947
00:15:26,000 --> 00:15:28,270
Okay, time for recap number three. Go
for it. Bullet one, RegVelo acts as a
948
00:15:28,270 --> 00:15:28,280
for it. Bullet one, RegVelo acts as a
949
00:15:28,280 --> 00:15:30,830
for it. Bullet one, RegVelo acts as a
virtual cell allowing in silico or
950
00:15:30,830 --> 00:15:30,840
virtual cell allowing in silico or
951
00:15:30,840 --> 00:15:33,150
virtual cell allowing in silico or
computer simulated gene knockouts.
952
00:15:33,150 --> 00:15:33,160
computer simulated gene knockouts.
953
00:15:33,160 --> 00:15:35,470
computer simulated gene knockouts.
Bullet two, it successfully mapped
954
00:15:35,470 --> 00:15:35,480
Bullet two, it successfully mapped
955
00:15:35,480 --> 00:15:37,830
Bullet two, it successfully mapped
pancreatic cell drivers like E2F1 and
956
00:15:37,830 --> 00:15:37,840
pancreatic cell drivers like E2F1 and
957
00:15:37,840 --> 00:15:38,750
pancreatic cell drivers like E2F1 and
RFY6.
958
00:15:38,750 --> 00:15:38,760
RFY6.
959
00:15:38,760 --> 00:15:39,270
RFY6.
>> Correct.
960
00:15:39,270 --> 00:15:39,280
>> Correct.
961
00:15:39,280 --> 00:15:41,430
>> Correct.
>> And bullet three, it perfectly simulated
962
00:15:41,430 --> 00:15:41,440
>> And bullet three, it perfectly simulated
963
00:15:41,440 --> 00:15:44,350
>> And bullet three, it perfectly simulated
the GATA1 versus SP1 toggle switch in
964
00:15:44,350 --> 00:15:44,360
the GATA1 versus SP1 toggle switch in
965
00:15:44,360 --> 00:15:46,550
the GATA1 versus SP1 toggle switch in
human blood cells. Awesome summary. But
966
00:15:46,550 --> 00:15:46,560
human blood cells. Awesome summary. But
967
00:15:46,560 --> 00:15:47,590
human blood cells. Awesome summary. But
you know, if you're hearing this and
968
00:15:47,590 --> 00:15:47,600
you know, if you're hearing this and
969
00:15:47,600 --> 00:15:49,590
you know, if you're hearing this and
thinking, "Okay, that's really cool. The
970
00:15:49,590 --> 00:15:49,600
thinking, "Okay, that's really cool. The
971
00:15:49,600 --> 00:15:51,630
thinking, "Okay, that's really cool. The
computer built a nice virtual cell that
972
00:15:51,630 --> 00:15:51,640
computer built a nice virtual cell that
973
00:15:51,640 --> 00:15:53,350
computer built a nice virtual cell that
confirms things we already discovered
974
00:15:53,350 --> 00:15:53,360
confirms things we already discovered
975
00:15:53,360 --> 00:15:54,470
confirms things we already discovered
years ago in textbooks."
976
00:15:54,470 --> 00:15:54,480
years ago in textbooks."
977
00:15:54,480 --> 00:15:55,870
years ago in textbooks."
>> Right, confirming the knowns.
978
00:15:55,870 --> 00:15:55,880
>> Right, confirming the knowns.
979
00:15:55,880 --> 00:15:58,110
>> Right, confirming the knowns.
>> Yeah. But does this algorithm actually
980
00:15:58,110 --> 00:15:58,120
>> Yeah. But does this algorithm actually
981
00:15:58,120 --> 00:16:00,590
>> Yeah. But does this algorithm actually
have the power to discover something
982
00:16:00,590 --> 00:16:00,600
have the power to discover something
983
00:16:00,600 --> 00:16:03,310
have the power to discover something
completely new? And that is exactly what
984
00:16:03,310 --> 00:16:03,320
completely new? And that is exactly what
985
00:16:03,320 --> 00:16:05,270
completely new? And that is exactly what
the researchers wanted to prove next.
986
00:16:05,270 --> 00:16:05,280
the researchers wanted to prove next.
987
00:16:05,280 --> 00:16:07,390
the researchers wanted to prove next.
Because validating known biology is one
988
00:16:07,390 --> 00:16:07,400
Because validating known biology is one
989
00:16:07,400 --> 00:16:09,790
Because validating known biology is one
thing. Predicting unknown biology is the
990
00:16:09,790 --> 00:16:09,800
thing. Predicting unknown biology is the
991
00:16:09,800 --> 00:16:11,870
thing. Predicting unknown biology is the
ultimate test of any computational
992
00:16:11,870 --> 00:16:11,880
ultimate test of any computational
993
00:16:11,880 --> 00:16:14,230
ultimate test of any computational
model. So to do this, the researchers
994
00:16:14,230 --> 00:16:14,240
model. So to do this, the researchers
995
00:16:14,240 --> 00:16:16,590
model. So to do this, the researchers
chose one of the most complex chaotic
996
00:16:16,590 --> 00:16:16,600
chose one of the most complex chaotic
997
00:16:16,600 --> 00:16:19,390
chose one of the most complex chaotic
systems in developmental biology. The
998
00:16:19,390 --> 00:16:19,400
systems in developmental biology. The
999
00:16:19,400 --> 00:16:21,790
systems in developmental biology. The
zebrafish cranial neural crest.
1000
00:16:21,790 --> 00:16:21,800
zebrafish cranial neural crest.
1001
00:16:21,800 --> 00:16:23,550
zebrafish cranial neural crest.
>> The neural crest. I mean, this is
1002
00:16:23,550 --> 00:16:23,560
>> The neural crest. I mean, this is
1003
00:16:23,560 --> 00:16:25,990
>> The neural crest. I mean, this is
Waddington's landscape on extreme fast
1004
00:16:25,990 --> 00:16:26,000
Waddington's landscape on extreme fast
1005
00:16:26,000 --> 00:16:26,590
Waddington's landscape on extreme fast
forward, isn't it?
1006
00:16:26,590 --> 00:16:26,600
forward, isn't it?
1007
00:16:26,600 --> 00:16:28,829
forward, isn't it?
>> Oh, it truly is. These are a temporary
1008
00:16:28,829 --> 00:16:28,839
>> Oh, it truly is. These are a temporary
1009
00:16:28,839 --> 00:16:31,110
>> Oh, it truly is. These are a temporary
group of cells in a developing embryo
1010
00:16:31,110 --> 00:16:31,120
group of cells in a developing embryo
1011
00:16:31,120 --> 00:16:33,550
group of cells in a developing embryo
that migrate incredibly fast and
1012
00:16:33,550 --> 00:16:33,560
that migrate incredibly fast and
1013
00:16:33,560 --> 00:16:35,790
that migrate incredibly fast and
differentiate into a wildly diverse
1014
00:16:35,790 --> 00:16:35,800
differentiate into a wildly diverse
1015
00:16:35,800 --> 00:16:37,950
differentiate into a wildly diverse
array of tissues. Like what? They turn
1016
00:16:37,950 --> 00:16:37,960
array of tissues. Like what? They turn
1017
00:16:37,960 --> 00:16:39,350
array of tissues. Like what? They turn
into the pigment cells that give the
1018
00:16:39,350 --> 00:16:39,360
into the pigment cells that give the
1019
00:16:39,360 --> 00:16:41,270
into the pigment cells that give the
fish its stripes, the cartilage in its
1020
00:16:41,270 --> 00:16:41,280
fish its stripes, the cartilage in its
1021
00:16:41,280 --> 00:16:44,070
fish its stripes, the cartilage in its
face, peripheral neurons, glia. It is a
1022
00:16:44,070 --> 00:16:44,080
face, peripheral neurons, glia. It is a
1023
00:16:44,080 --> 00:16:47,750
face, peripheral neurons, glia. It is a
very noisy, highly dynamic system. So to
1024
00:16:47,750 --> 00:16:47,760
very noisy, highly dynamic system. So to
1025
00:16:47,760 --> 00:16:49,590
very noisy, highly dynamic system. So to
give the algorithm the best possible
1026
00:16:49,590 --> 00:16:49,600
give the algorithm the best possible
1027
00:16:49,600 --> 00:16:51,510
give the algorithm the best possible
chance at mapping this chaos, the
1028
00:16:51,510 --> 00:16:51,520
chance at mapping this chaos, the
1029
00:16:51,520 --> 00:16:53,030
chance at mapping this chaos, the
researchers didn't just use standard
1030
00:16:53,030 --> 00:16:53,040
researchers didn't just use standard
1031
00:16:53,040 --> 00:16:54,910
researchers didn't just use standard
sequencing. No, they couldn't.
1032
00:16:54,910 --> 00:16:54,920
sequencing. No, they couldn't.
1033
00:16:54,920 --> 00:16:56,750
sequencing. No, they couldn't.
>> They threw the absolute state of the art
1034
00:16:56,750 --> 00:16:56,760
>> They threw the absolute state of the art
1035
00:16:56,760 --> 00:16:59,310
>> They threw the absolute state of the art
at it, generating a massive multiomic
1036
00:16:59,310 --> 00:16:59,320
at it, generating a massive multiomic
1037
00:16:59,320 --> 00:17:01,470
at it, generating a massive multiomic
data stack. They needed extremely
1038
00:17:01,470 --> 00:17:01,480
data stack. They needed extremely
1039
00:17:01,480 --> 00:17:03,190
data stack. They needed extremely
high-fidelity input. So they used a
1040
00:17:03,190 --> 00:17:03,200
high-fidelity input. So they used a
1041
00:17:03,200 --> 00:17:06,150
high-fidelity input. So they used a
technique called full-length Smart-seq3
1042
00:17:06,150 --> 00:17:06,160
technique called full-length Smart-seq3
1043
00:17:06,160 --> 00:17:08,670
technique called full-length Smart-seq3
to get incredibly high-resolution RNA
1044
00:17:08,670 --> 00:17:08,680
to get incredibly high-resolution RNA
1045
00:17:08,680 --> 00:17:11,590
to get incredibly high-resolution RNA
data. Smart-seq3? Yes. But more
1046
00:17:11,590 --> 00:17:11,600
data. Smart-seq3? Yes. But more
1047
00:17:11,600 --> 00:17:13,590
data. Smart-seq3? Yes. But more
importantly, they measured that RNA
1048
00:17:13,590 --> 00:17:13,600
importantly, they measured that RNA
1049
00:17:13,600 --> 00:17:15,670
importantly, they measured that RNA
expression while simultaneously
1050
00:17:15,670 --> 00:17:15,680
expression while simultaneously
1051
00:17:15,680 --> 00:17:18,030
expression while simultaneously
measuring the chromatin accessibility.
1052
00:17:18,030 --> 00:17:18,040
measuring the chromatin accessibility.
1053
00:17:18,040 --> 00:17:20,710
measuring the chromatin accessibility.
Meaning how physically open the DNA was.
1054
00:17:20,710 --> 00:17:20,720
Meaning how physically open the DNA was.
1055
00:17:20,720 --> 00:17:22,790
Meaning how physically open the DNA was.
>> Exactly. In the exact same individual
1056
00:17:22,790 --> 00:17:22,800
>> Exactly. In the exact same individual
1057
00:17:22,800 --> 00:17:24,790
>> Exactly. In the exact same individual
cells. So, they were essentially reading
1058
00:17:24,790 --> 00:17:24,800
cells. So, they were essentially reading
1059
00:17:24,800 --> 00:17:27,230
cells. So, they were essentially reading
the text of the instruction manual while
1060
00:17:27,230 --> 00:17:27,240
the text of the instruction manual while
1061
00:17:27,240 --> 00:17:29,230
the text of the instruction manual while
simultaneously checking the spine of the
1062
00:17:29,230 --> 00:17:29,240
simultaneously checking the spine of the
1063
00:17:29,240 --> 00:17:31,150
simultaneously checking the spine of the
book to see which pages naturally fell
1064
00:17:31,150 --> 00:17:31,160
book to see which pages naturally fell
1065
00:17:31,160 --> 00:17:32,870
book to see which pages naturally fell
open. That is a brilliant way to picture
1066
00:17:32,870 --> 00:17:32,880
open. That is a brilliant way to picture
1067
00:17:32,880 --> 00:17:35,150
open. That is a brilliant way to picture
it, and it gave RegVelo an unbelievably
1068
00:17:35,150 --> 00:17:35,160
it, and it gave RegVelo an unbelievably
1069
00:17:35,160 --> 00:17:36,350
it, and it gave RegVelo an unbelievably
rich starting map.
1070
00:17:36,350 --> 00:17:36,360
rich starting map.
1071
00:17:36,360 --> 00:17:38,190
rich starting map.
>> So, they fed all of this high-resolution
1072
00:17:38,190 --> 00:17:38,200
>> So, they fed all of this high-resolution
1073
00:17:38,200 --> 00:17:40,190
>> So, they fed all of this high-resolution
data into the framework. Right. And
1074
00:17:40,190 --> 00:17:40,200
data into the framework. Right. And
1075
00:17:40,200 --> 00:17:42,230
data into the framework. Right. And
RegVelo's algorithms crunched those
1076
00:17:42,230 --> 00:17:42,240
RegVelo's algorithms crunched those
1077
00:17:42,240 --> 00:17:44,110
RegVelo's algorithms crunched those
high-dimensional equations, calculated
1078
00:17:44,110 --> 00:17:44,120
high-dimensional equations, calculated
1079
00:17:44,120 --> 00:17:45,790
high-dimensional equations, calculated
the latent times, self-corrected the
1080
00:17:45,790 --> 00:17:45,800
the latent times, self-corrected the
1081
00:17:45,800 --> 00:17:47,630
the latent times, self-corrected the
network weights. Using sparsity, like we
1082
00:17:47,630 --> 00:17:47,640
network weights. Using sparsity, like we
1083
00:17:47,640 --> 00:17:48,230
network weights. Using sparsity, like we
talked about.
1084
00:17:48,230 --> 00:17:48,240
talked about.
1085
00:17:48,240 --> 00:17:50,670
talked about.
>> Yes. And it output a prediction. It
1086
00:17:50,670 --> 00:17:50,680
>> Yes. And it output a prediction. It
1087
00:17:50,680 --> 00:17:52,830
>> Yes. And it output a prediction. It
pointed to specific very early
1088
00:17:52,830 --> 00:17:52,840
pointed to specific very early
1089
00:17:52,840 --> 00:17:54,870
pointed to specific very early
transcription factors as the master
1090
00:17:54,870 --> 00:17:54,880
transcription factors as the master
1091
00:17:54,880 --> 00:17:56,910
transcription factors as the master
switches for pigment cell development.
1092
00:17:56,910 --> 00:17:56,920
switches for pigment cell development.
1093
00:17:56,920 --> 00:18:00,230
switches for pigment cell development.
>> Specifically, it flagged two genes, Tfec
1094
00:18:00,230 --> 00:18:00,240
>> Specifically, it flagged two genes, Tfec
1095
00:18:00,240 --> 00:18:02,430
>> Specifically, it flagged two genes, Tfec
and Elf1. Now, scientists knew these
1096
00:18:02,430 --> 00:18:02,440
and Elf1. Now, scientists knew these
1097
00:18:02,440 --> 00:18:03,750
and Elf1. Now, scientists knew these
genes were involved in development
1098
00:18:03,750 --> 00:18:03,760
genes were involved in development
1099
00:18:03,760 --> 00:18:05,710
genes were involved in development
somewhere. Right. But their role as the
1100
00:18:05,710 --> 00:18:05,720
somewhere. Right. But their role as the
1101
00:18:05,720 --> 00:18:08,230
somewhere. Right. But their role as the
primary earliest drivers dictating the
1102
00:18:08,230 --> 00:18:08,240
primary earliest drivers dictating the
1103
00:18:08,240 --> 00:18:09,870
primary earliest drivers dictating the
pigment lineage was heavily
1104
00:18:09,870 --> 00:18:09,880
pigment lineage was heavily
1105
00:18:09,880 --> 00:18:12,150
pigment lineage was heavily
underappreciated. RegVelo basically
1106
00:18:12,150 --> 00:18:12,160
underappreciated. RegVelo basically
1107
00:18:12,160 --> 00:18:14,550
underappreciated. RegVelo basically
looked at the math and said, "These two
1108
00:18:14,550 --> 00:18:14,560
looked at the math and said, "These two
1109
00:18:14,560 --> 00:18:16,790
looked at the math and said, "These two
genes are the most critical traffic cops
1110
00:18:16,790 --> 00:18:16,800
genes are the most critical traffic cops
1111
00:18:16,800 --> 00:18:19,070
genes are the most critical traffic cops
at the very top of the hill." And this
1112
00:18:19,070 --> 00:18:19,080
at the very top of the hill." And this
1113
00:18:19,080 --> 00:18:20,430
at the very top of the hill." And this
is honestly my favorite part of the
1114
00:18:20,430 --> 00:18:20,440
is honestly my favorite part of the
1115
00:18:20,440 --> 00:18:21,230
is honestly my favorite part of the
entire paper.
1116
00:18:21,230 --> 00:18:21,240
entire paper.
1117
00:18:21,240 --> 00:18:23,790
entire paper.
>> Oh, the validation. Yes.
1118
00:18:23,790 --> 00:18:23,800
>> Oh, the validation. Yes.
1119
00:18:23,800 --> 00:18:25,270
>> Oh, the validation. Yes.
The researchers didn't just publish
1120
00:18:25,270 --> 00:18:25,280
The researchers didn't just publish
1121
00:18:25,280 --> 00:18:26,990
The researchers didn't just publish
their neat little computer simulation
1122
00:18:26,990 --> 00:18:27,000
their neat little computer simulation
1123
00:18:27,000 --> 00:18:28,790
their neat little computer simulation
and walk away. No, they didn't.
1124
00:18:28,790 --> 00:18:28,800
and walk away. No, they didn't.
1125
00:18:28,800 --> 00:18:30,470
and walk away. No, they didn't.
>> They took RegVelo's mathematical
1126
00:18:30,470 --> 00:18:30,480
>> They took RegVelo's mathematical
1127
00:18:30,480 --> 00:18:32,710
>> They took RegVelo's mathematical
prediction and marched straight into the
1128
00:18:32,710 --> 00:18:32,720
prediction and marched straight into the
1129
00:18:32,720 --> 00:18:34,870
prediction and marched straight into the
physical wet lab. They performed what we
1130
00:18:34,870 --> 00:18:34,880
physical wet lab. They performed what we
1131
00:18:34,880 --> 00:18:37,990
physical wet lab. They performed what we
call in vivo knockout validation. They
1132
00:18:37,990 --> 00:18:38,000
call in vivo knockout validation. They
1133
00:18:38,000 --> 00:18:40,230
call in vivo knockout validation. They
tested it in living animals.
1134
00:18:40,230 --> 00:18:40,240
tested it in living animals.
1135
00:18:40,240 --> 00:18:42,910
tested it in living animals.
>> They took actual living zebrafish
1136
00:18:42,910 --> 00:18:42,920
>> They took actual living zebrafish
1137
00:18:42,920 --> 00:18:45,830
>> They took actual living zebrafish
embryos. Yes. They used CRISPR Cas9
1138
00:18:45,830 --> 00:18:45,840
embryos. Yes. They used CRISPR Cas9
1139
00:18:45,840 --> 00:18:48,350
embryos. Yes. They used CRISPR Cas9
genetic scissors to physically snip the
1140
00:18:48,350 --> 00:18:48,360
genetic scissors to physically snip the
1141
00:18:48,360 --> 00:18:50,990
genetic scissors to physically snip the
Tfec and Elf1 genes right out of the
1142
00:18:50,990 --> 00:18:51,000
Tfec and Elf1 genes right out of the
1143
00:18:51,000 --> 00:18:53,270
Tfec and Elf1 genes right out of the
DNA. And then they let the embryos grow
1144
00:18:53,270 --> 00:18:53,280
DNA. And then they let the embryos grow
1145
00:18:53,280 --> 00:18:54,830
DNA. And then they let the embryos grow
and used advanced single-cell
1146
00:18:54,830 --> 00:18:54,840
and used advanced single-cell
1147
00:18:54,840 --> 00:18:57,430
and used advanced single-cell
sequencing, specifically Perturb-seq, to
1148
00:18:57,430 --> 00:18:57,440
sequencing, specifically Perturb-seq, to
1149
00:18:57,440 --> 00:18:59,110
sequencing, specifically Perturb-seq, to
measure exactly what happened to those
1150
00:18:59,110 --> 00:18:59,120
measure exactly what happened to those
1151
00:18:59,120 --> 00:19:01,230
measure exactly what happened to those
physical cells. And the results were
1152
00:19:01,230 --> 00:19:01,240
physical cells. And the results were
1153
00:19:01,240 --> 00:19:03,390
physical cells. And the results were
undeniable. The physical biological
1154
00:19:03,390 --> 00:19:03,400
undeniable. The physical biological
1155
00:19:03,400 --> 00:19:05,990
undeniable. The physical biological
knockout perfectly validated the digital
1156
00:19:05,990 --> 00:19:06,000
knockout perfectly validated the digital
1157
00:19:06,000 --> 00:19:08,710
knockout perfectly validated the digital
mathematical simulation. When Tfec and
1158
00:19:08,710 --> 00:19:08,720
mathematical simulation. When Tfec and
1159
00:19:08,720 --> 00:19:11,270
mathematical simulation. When Tfec and
Elf1 were removed from in living fish,
1160
00:19:11,270 --> 00:19:11,280
Elf1 were removed from in living fish,
1161
00:19:11,280 --> 00:19:13,190
Elf1 were removed from in living fish,
the pigment lineage specification
1162
00:19:13,190 --> 00:19:13,200
the pigment lineage specification
1163
00:19:13,200 --> 00:19:15,590
the pigment lineage specification
completely broke down. The cells
1164
00:19:15,590 --> 00:19:15,600
completely broke down. The cells
1165
00:19:15,600 --> 00:19:17,670
completely broke down. The cells
entirely failed to differentiate into
1166
00:19:17,670 --> 00:19:17,680
entirely failed to differentiate into
1167
00:19:17,680 --> 00:19:19,750
entirely failed to differentiate into
pigment cells, matching RegVelo's
1168
00:19:19,750 --> 00:19:19,760
pigment cells, matching RegVelo's
1169
00:19:19,760 --> 00:19:22,030
pigment cells, matching RegVelo's
virtual prediction flawlessly. The
1170
00:19:22,030 --> 00:19:22,040
virtual prediction flawlessly. The
1171
00:19:22,040 --> 00:19:24,710
virtual prediction flawlessly. The
algorithm nailed it. The virtual model
1172
00:19:24,710 --> 00:19:24,720
algorithm nailed it. The virtual model
1173
00:19:24,720 --> 00:19:27,150
algorithm nailed it. The virtual model
was so tightly coupled to reality that
1174
00:19:27,150 --> 00:19:27,160
was so tightly coupled to reality that
1175
00:19:27,160 --> 00:19:29,030
was so tightly coupled to reality that
it predicted a physical outcome in a
1176
00:19:29,030 --> 00:19:29,040
it predicted a physical outcome in a
1177
00:19:29,040 --> 00:19:31,470
it predicted a physical outcome in a
living organism. And the statistics back
1178
00:19:31,470 --> 00:19:31,480
living organism. And the statistics back
1179
00:19:31,480 --> 00:19:34,390
living organism. And the statistics back
up just how significant a leap this is.
1180
00:19:34,390 --> 00:19:34,400
up just how significant a leap this is.
1181
00:19:34,400 --> 00:19:36,270
up just how significant a leap this is.
When the team measured how accurately
1182
00:19:36,270 --> 00:19:36,280
When the team measured how accurately
1183
00:19:36,280 --> 00:19:38,430
When the team measured how accurately
RegVelo predicted the downstream effects
1184
00:19:38,430 --> 00:19:38,440
RegVelo predicted the downstream effects
1185
00:19:38,440 --> 00:19:39,830
RegVelo predicted the downstream effects
of knocking out these individual
1186
00:19:39,830 --> 00:19:39,840
of knocking out these individual
1187
00:19:39,840 --> 00:19:42,390
of knocking out these individual
transcription factors, RegVelo achieved
1188
00:19:42,390 --> 00:19:42,400
transcription factors, RegVelo achieved
1189
00:19:42,400 --> 00:19:45,830
transcription factors, RegVelo achieved
a mean Spearman correlation of 0.52.
1190
00:19:45,830 --> 00:19:45,840
a mean Spearman correlation of 0.52.
1191
00:19:45,840 --> 00:19:47,230
a mean Spearman correlation of 0.52.
Okay, I know a Spearman correlation of
1192
00:19:47,230 --> 00:19:47,240
Okay, I know a Spearman correlation of
1193
00:19:47,240 --> 00:19:49,510
Okay, I know a Spearman correlation of
0.52 sounds like deep statistical
1194
00:19:49,510 --> 00:19:49,520
0.52 sounds like deep statistical
1195
00:19:49,520 --> 00:19:51,270
0.52 sounds like deep statistical
jargon. It is a bit dense, yes.
1196
00:19:51,270 --> 00:19:51,280
jargon. It is a bit dense, yes.
1197
00:19:51,280 --> 00:19:52,990
jargon. It is a bit dense, yes.
>> But let's put that number in context for
1198
00:19:52,990 --> 00:19:53,000
>> But let's put that number in context for
1199
00:19:53,000 --> 00:19:54,870
>> But let's put that number in context for
the listener. Compared to the older
1200
00:19:54,870 --> 00:19:54,880
the listener. Compared to the older
1201
00:19:54,880 --> 00:19:57,230
the listener. Compared to the older
decoupled RNA velocity methods we talked
1202
00:19:57,230 --> 00:19:57,240
decoupled RNA velocity methods we talked
1203
00:19:57,240 --> 00:19:58,790
decoupled RNA velocity methods we talked
about at the beginning of the show, how
1204
00:19:58,790 --> 00:19:58,800
about at the beginning of the show, how
1205
00:19:58,800 --> 00:20:01,070
about at the beginning of the show, how
much better is that 0.52? It is nearly
1206
00:20:01,070 --> 00:20:01,080
much better is that 0.52? It is nearly
1207
00:20:01,080 --> 00:20:02,190
much better is that 0.52? It is nearly
double the accuracy.
1208
00:20:02,190 --> 00:20:02,200
double the accuracy.
1209
00:20:02,200 --> 00:20:03,670
double the accuracy.
>> Double? Yes.
1210
00:20:03,670 --> 00:20:03,680
>> Double? Yes.
1211
00:20:03,680 --> 00:20:05,030
>> Double? Yes.
By connecting the gene regulatory
1212
00:20:05,030 --> 00:20:05,040
By connecting the gene regulatory
1213
00:20:05,040 --> 00:20:06,790
By connecting the gene regulatory
network to the velocity dynamics, they
1214
00:20:06,790 --> 00:20:06,800
network to the velocity dynamics, they
1215
00:20:06,800 --> 00:20:08,670
network to the velocity dynamics, they
doubled their ability to predict complex
1216
00:20:08,670 --> 00:20:08,680
doubled their ability to predict complex
1217
00:20:08,680 --> 00:20:11,670
doubled their ability to predict complex
biological reality. That is huge. So,
1218
00:20:11,670 --> 00:20:11,680
biological reality. That is huge. So,
1219
00:20:11,680 --> 00:20:13,350
biological reality. That is huge. So,
let's do our final mini recap. Bullet
1220
00:20:13,350 --> 00:20:13,360
let's do our final mini recap. Bullet
1221
00:20:13,360 --> 00:20:15,790
let's do our final mini recap. Bullet
one, RegVelo was tested on the extremely
1222
00:20:15,790 --> 00:20:15,800
one, RegVelo was tested on the extremely
1223
00:20:15,800 --> 00:20:18,550
one, RegVelo was tested on the extremely
complex zebrafish cranial neural crest.
1224
00:20:18,550 --> 00:20:18,560
complex zebrafish cranial neural crest.
1225
00:20:18,560 --> 00:20:21,470
complex zebrafish cranial neural crest.
Yep. Bullet two, it predicted TFEB and
1226
00:20:21,470 --> 00:20:21,480
Yep. Bullet two, it predicted TFEB and
1227
00:20:21,480 --> 00:20:24,110
Yep. Bullet two, it predicted TFEB and
Elf1 as the crucial drivers for pigment
1228
00:20:24,110 --> 00:20:24,120
Elf1 as the crucial drivers for pigment
1229
00:20:24,120 --> 00:20:24,550
Elf1 as the crucial drivers for pigment
cells.
1230
00:20:24,550 --> 00:20:24,560
cells.
1231
00:20:24,560 --> 00:20:27,270
cells.
>> Right. And bullet three, CRISPR-Cas9
1232
00:20:27,270 --> 00:20:27,280
>> Right. And bullet three, CRISPR-Cas9
1233
00:20:27,280 --> 00:20:29,510
>> Right. And bullet three, CRISPR-Cas9
knockouts in live embryos proved the
1234
00:20:29,510 --> 00:20:29,520
knockouts in live embryos proved the
1235
00:20:29,520 --> 00:20:32,030
knockouts in live embryos proved the
computer model was right. Spot on. So,
1236
00:20:32,030 --> 00:20:32,040
computer model was right. Spot on. So,
1237
00:20:32,040 --> 00:20:33,990
computer model was right. Spot on. So,
the question to ask yourself now is,
1238
00:20:33,990 --> 00:20:34,000
the question to ask yourself now is,
1239
00:20:34,000 --> 00:20:35,670
the question to ask yourself now is,
what does this actually mean for the
1240
00:20:35,670 --> 00:20:35,680
what does this actually mean for the
1241
00:20:35,680 --> 00:20:37,750
what does this actually mean for the
future of how we study diseases and
1242
00:20:37,750 --> 00:20:37,760
future of how we study diseases and
1243
00:20:37,760 --> 00:20:39,910
future of how we study diseases and
developmental biology? Well, it means
1244
00:20:39,910 --> 00:20:39,920
developmental biology? Well, it means
1245
00:20:39,920 --> 00:20:42,510
developmental biology? Well, it means
we're moving from simply drawing lines
1246
00:20:42,510 --> 00:20:42,520
we're moving from simply drawing lines
1247
00:20:42,520 --> 00:20:45,030
we're moving from simply drawing lines
on a graph and observing a car drive by
1248
00:20:45,030 --> 00:20:45,040
on a graph and observing a car drive by
1249
00:20:45,040 --> 00:20:46,870
on a graph and observing a car drive by
to actually understanding how the engine
1250
00:20:46,870 --> 00:20:46,880
to actually understanding how the engine
1251
00:20:46,880 --> 00:20:48,830
to actually understanding how the engine
turns the wheels. It represents a
1252
00:20:48,830 --> 00:20:48,840
turns the wheels. It represents a
1253
00:20:48,840 --> 00:20:50,590
turns the wheels. It represents a
fundamental paradigm shift in
1254
00:20:50,590 --> 00:20:50,600
fundamental paradigm shift in
1255
00:20:50,600 --> 00:20:52,870
fundamental paradigm shift in
computational biology. It does. We are
1256
00:20:52,870 --> 00:20:52,880
computational biology. It does. We are
1257
00:20:52,880 --> 00:20:54,350
computational biology. It does. We are
transitioning from an observational
1258
00:20:54,350 --> 00:20:54,360
transitioning from an observational
1259
00:20:54,360 --> 00:20:56,790
transitioning from an observational
phase into a causal phase. By honoring
1260
00:20:56,790 --> 00:20:56,800
phase into a causal phase. By honoring
1261
00:20:56,800 --> 00:20:58,870
phase into a causal phase. By honoring
the fact that biology is highly coupled
1262
00:20:58,870 --> 00:20:58,880
the fact that biology is highly coupled
1263
00:20:58,880 --> 00:21:00,830
the fact that biology is highly coupled
and interactive, RegVelo provides an
1264
00:21:00,830 --> 00:21:00,840
and interactive, RegVelo provides an
1265
00:21:00,840 --> 00:21:03,430
and interactive, RegVelo provides an
interpretable, actionable framework. We
1266
00:21:03,430 --> 00:21:03,440
interpretable, actionable framework. We
1267
00:21:03,440 --> 00:21:05,270
interpretable, actionable framework. We
can now test hypotheses in a computer
1268
00:21:05,270 --> 00:21:05,280
can now test hypotheses in a computer
1269
00:21:05,280 --> 00:21:07,470
can now test hypotheses in a computer
simulation with a level of mechanical
1270
00:21:07,470 --> 00:21:07,480
simulation with a level of mechanical
1271
00:21:07,480 --> 00:21:09,390
simulation with a level of mechanical
confidence that simply did not exist
1272
00:21:09,390 --> 00:21:09,400
confidence that simply did not exist
1273
00:21:09,400 --> 00:21:11,150
confidence that simply did not exist
before this paper. And that leads to a
1274
00:21:11,150 --> 00:21:11,160
before this paper. And that leads to a
1275
00:21:11,160 --> 00:21:12,430
before this paper. And that leads to a
thought I really want to leave you with
1276
00:21:12,430 --> 00:21:12,440
thought I really want to leave you with
1277
00:21:12,440 --> 00:21:14,390
thought I really want to leave you with
as we wrapped up today's deep dive.
1278
00:21:14,390 --> 00:21:14,400
as we wrapped up today's deep dive.
1279
00:21:14,400 --> 00:21:15,990
as we wrapped up today's deep dive.
Think about the implications of this for
1280
00:21:15,990 --> 00:21:16,000
Think about the implications of this for
1281
00:21:16,000 --> 00:21:17,270
Think about the implications of this for
medicine.
1282
00:21:17,270 --> 00:21:17,280
medicine.
1283
00:21:17,280 --> 00:21:18,630
medicine.
We just spent a lot of time talking
1284
00:21:18,630 --> 00:21:18,640
We just spent a lot of time talking
1285
00:21:18,640 --> 00:21:20,550
We just spent a lot of time talking
about predicting how a healthy stem cell
1286
00:21:20,550 --> 00:21:20,560
about predicting how a healthy stem cell
1287
00:21:20,560 --> 00:21:22,510
about predicting how a healthy stem cell
becomes a blood cell or how it becomes a
1288
00:21:22,510 --> 00:21:22,520
becomes a blood cell or how it becomes a
1289
00:21:22,520 --> 00:21:24,790
becomes a blood cell or how it becomes a
pigment cell in a fish. Mhm. But
1290
00:21:24,790 --> 00:21:24,800
pigment cell in a fish. Mhm. But
1291
00:21:24,800 --> 00:21:26,670
pigment cell in a fish. Mhm. But
Waddington's landscape isn't just about
1292
00:21:26,670 --> 00:21:26,680
Waddington's landscape isn't just about
1293
00:21:26,680 --> 00:21:28,630
Waddington's landscape isn't just about
healthy development. Cancer is
1294
00:21:28,630 --> 00:21:28,640
healthy development. Cancer is
1295
00:21:28,640 --> 00:21:30,670
healthy development. Cancer is
essentially a problem of cell fate and
1296
00:21:30,670 --> 00:21:30,680
essentially a problem of cell fate and
1297
00:21:30,680 --> 00:21:33,470
essentially a problem of cell fate and
gene regulation gone horribly tragically
1298
00:21:33,470 --> 00:21:33,480
gene regulation gone horribly tragically
1299
00:21:33,480 --> 00:21:35,710
gene regulation gone horribly tragically
wrong. Exactly. It is a mutated
1300
00:21:35,710 --> 00:21:35,720
wrong. Exactly. It is a mutated
1301
00:21:35,720 --> 00:21:37,710
wrong. Exactly. It is a mutated
landscape where the marble gets trapped
1302
00:21:37,710 --> 00:21:37,720
landscape where the marble gets trapped
1303
00:21:37,720 --> 00:21:39,270
landscape where the marble gets trapped
in a valley it was never supposed to
1304
00:21:39,270 --> 00:21:39,280
in a valley it was never supposed to
1305
00:21:39,280 --> 00:21:41,710
in a valley it was never supposed to
enter and it just keeps dividing. So if
1306
00:21:41,710 --> 00:21:41,720
enter and it just keeps dividing. So if
1307
00:21:41,720 --> 00:21:44,150
enter and it just keeps dividing. So if
we now have the computational power to
1308
00:21:44,150 --> 00:21:44,160
we now have the computational power to
1309
00:21:44,160 --> 00:21:47,350
we now have the computational power to
accurately simulate a virtual cell and
1310
00:21:47,350 --> 00:21:47,360
accurately simulate a virtual cell and
1311
00:21:47,360 --> 00:21:49,470
accurately simulate a virtual cell and
predict exactly how altering its wiring
1312
00:21:49,470 --> 00:21:49,480
predict exactly how altering its wiring
1313
00:21:49,480 --> 00:21:50,470
predict exactly how altering its wiring
changes its fate.
1314
00:21:50,470 --> 00:21:50,480
changes its fate.
1315
00:21:50,480 --> 00:21:52,150
changes its fate.
>> Where does that leave us? Right. How far
1316
00:21:52,150 --> 00:21:52,160
>> Where does that leave us? Right. How far
1317
00:21:52,160 --> 00:21:54,190
>> Where does that leave us? Right. How far
are we from a world where a doctor
1318
00:21:54,190 --> 00:21:54,200
are we from a world where a doctor
1319
00:21:54,200 --> 00:21:56,750
are we from a world where a doctor
biopsies your specific tumor, they
1320
00:21:56,750 --> 00:21:56,760
biopsies your specific tumor, they
1321
00:21:56,760 --> 00:21:58,790
biopsies your specific tumor, they
sequence the RNA, they measure the DNA
1322
00:21:58,790 --> 00:21:58,800
sequence the RNA, they measure the DNA
1323
00:21:58,800 --> 00:22:00,950
sequence the RNA, they measure the DNA
accessibility and they upload your exact
1324
00:22:00,950 --> 00:22:00,960
accessibility and they upload your exact
1325
00:22:00,960 --> 00:22:02,990
accessibility and they upload your exact
cancer cells into a framework like
1326
00:22:02,990 --> 00:22:03,000
cancer cells into a framework like
1327
00:22:03,000 --> 00:22:03,550
cancer cells into a framework like
RigVelo.
1328
00:22:03,550 --> 00:22:03,560
RigVelo.
1329
00:22:03,560 --> 00:22:04,550
RigVelo.
>> They could create a completely
1330
00:22:04,550 --> 00:22:04,560
>> They could create a completely
1331
00:22:04,560 --> 00:22:07,790
>> They could create a completely
personalized digital twin of your tumor.
1332
00:22:07,790 --> 00:22:07,800
personalized digital twin of your tumor.
1333
00:22:07,800 --> 00:22:10,150
personalized digital twin of your tumor.
Yes. And instead of testing highly toxic
1334
00:22:10,150 --> 00:22:10,160
Yes. And instead of testing highly toxic
1335
00:22:10,160 --> 00:22:12,550
Yes. And instead of testing highly toxic
chemotherapy drugs on your actual body
1336
00:22:12,550 --> 00:22:12,560
chemotherapy drugs on your actual body
1337
00:22:12,560 --> 00:22:14,710
chemotherapy drugs on your actual body
through months of agonizing trial and
1338
00:22:14,710 --> 00:22:14,720
through months of agonizing trial and
1339
00:22:14,720 --> 00:22:16,670
through months of agonizing trial and
error, the doctor performs in silico
1340
00:22:16,670 --> 00:22:16,680
error, the doctor performs in silico
1341
00:22:16,680 --> 00:22:19,350
error, the doctor performs in silico
perturbations. Exactly. They digitally
1342
00:22:19,350 --> 00:22:19,360
perturbations. Exactly. They digitally
1343
00:22:19,360 --> 00:22:21,190
perturbations. Exactly. They digitally
simulate thousands of drug combinations
1344
00:22:21,190 --> 00:22:21,200
simulate thousands of drug combinations
1345
00:22:21,200 --> 00:22:22,990
simulate thousands of drug combinations
in the computer knocking out different
1346
00:22:22,990 --> 00:22:23,000
in the computer knocking out different
1347
00:22:23,000 --> 00:22:25,110
in the computer knocking out different
transcription factors until they find
1348
00:22:25,110 --> 00:22:25,120
transcription factors until they find
1349
00:22:25,120 --> 00:22:28,110
transcription factors until they find
the exact molecular mechanism that
1350
00:22:28,110 --> 00:22:28,120
the exact molecular mechanism that
1351
00:22:28,120 --> 00:22:30,430
the exact molecular mechanism that
forces your specific cancer cells to
1352
00:22:30,430 --> 00:22:30,440
forces your specific cancer cells to
1353
00:22:30,440 --> 00:22:32,470
forces your specific cancer cells to
stop growing. Or forces them to
1354
00:22:32,470 --> 00:22:32,480
stop growing. Or forces them to
1355
00:22:32,480 --> 00:22:34,390
stop growing. Or forces them to
differentiate into harmless tissue.
1356
00:22:34,390 --> 00:22:34,400
differentiate into harmless tissue.
1357
00:22:34,400 --> 00:22:36,950
differentiate into harmless tissue.
>> Yes. They solve the puzzle virtually
1358
00:22:36,950 --> 00:22:36,960
>> Yes. They solve the puzzle virtually
1359
00:22:36,960 --> 00:22:38,710
>> Yes. They solve the puzzle virtually
before you ever have to swallow a single
1360
00:22:38,710 --> 00:22:38,720
before you ever have to swallow a single
1361
00:22:38,720 --> 00:22:40,950
before you ever have to swallow a single
pill. That is the ultimate promise of
1362
00:22:40,950 --> 00:22:40,960
pill. That is the ultimate promise of
1363
00:22:40,960 --> 00:22:42,510
pill. That is the ultimate promise of
this technology.
1364
00:22:42,510 --> 00:22:42,520
this technology.
1365
00:22:42,520 --> 00:22:44,550
this technology.
By finally linking the genetic software
1366
00:22:44,550 --> 00:22:44,560
By finally linking the genetic software
1367
00:22:44,560 --> 00:22:46,870
By finally linking the genetic software
to the cellular dynamics, we aren't just
1368
00:22:46,870 --> 00:22:46,880
to the cellular dynamics, we aren't just
1369
00:22:46,880 --> 00:22:49,150
to the cellular dynamics, we aren't just
observing biology anymore. We are
1370
00:22:49,150 --> 00:22:49,160
observing biology anymore. We are
1371
00:22:49,160 --> 00:22:51,430
observing biology anymore. We are
learning how to actively rewrite it. So
1372
00:22:51,430 --> 00:22:51,440
learning how to actively rewrite it. So
1373
00:22:51,440 --> 00:22:53,030
learning how to actively rewrite it. So
the next time you picture that single
1374
00:22:53,030 --> 00:22:53,040
the next time you picture that single
1375
00:22:53,040 --> 00:22:55,070
the next time you picture that single
marble rolling down the bumpy slopes of
1376
00:22:55,070 --> 00:22:55,080
marble rolling down the bumpy slopes of
1377
00:22:55,080 --> 00:22:56,870
marble rolling down the bumpy slopes of
Waddington's landscape bouncing between
1378
00:22:56,870 --> 00:22:56,880
Waddington's landscape bouncing between
1379
00:22:56,880 --> 00:22:59,070
Waddington's landscape bouncing between
the ridges and valleys, remember that we
1380
00:22:59,070 --> 00:22:59,080
the ridges and valleys, remember that we
1381
00:22:59,080 --> 00:23:01,310
the ridges and valleys, remember that we
are no longer just watching it roll. No,
1382
00:23:01,310 --> 00:23:01,320
are no longer just watching it roll. No,
1383
00:23:01,320 --> 00:23:03,230
are no longer just watching it roll. No,
we're not. We finally have the tools to
1384
00:23:03,230 --> 00:23:03,240
we're not. We finally have the tools to
1385
00:23:03,240 --> 00:23:05,230
we're not. We finally have the tools to
read the map and maybe very soon we'll
1386
00:23:05,230 --> 00:23:05,240
read the map and maybe very soon we'll
1387
00:23:05,240 --> 00:23:08,680
read the map and maybe very soon we'll
be the ones drawing the valleys.
125090
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.
Afrikaans
Akan
Albanian
Amharic
Arabic
Armenian
Azerbaijani
Basque
Belarusian
Bemba
Bengali
Bihari
Bosnian
Breton
Bulgarian
Cambodian
Catalan
Cebuano
Cherokee
Chichewa
Chinese (Traditional)
Corsican
Croatian
Czech
Danish
Dutch
English
Esperanto
Estonian
Faroese
Finnish
Galician
Georgian
German
Greek
Guarani
Haitian Creole
Hausa
Hawaiian
Hebrew
Hmong
Hungarian
Icelandic
Indonesian
Irish
Italian
Japanese
Kazakh
Kinyarwanda
Kirundi
Kongo
Korean
Krio (Sierra Leone)
Kurdish
Kyrgyz
Laothian
Latin
Latvian
Lithuanian
Luganda
Luo
Luxembourgish
Macedonian
Malagasy
Malay
Maltese
Maori
Mauritian Creole
Moldavian
Mongolian
Myanmar (Burmese)
Montenegrin
Nepali
Norwegian
Pashto
Persian
Polish
Portuguese (Brazil)
Portuguese (Portugal)
Quechua
Romanian
Romansh
Russian
Samoan
Scots Gaelic
Serbian
Sesotho
Setswana
Seychellois Creole
Shona
Sindhi
Sinhalese
Slovak
Slovenian
Somali
Spanish
Swedish
Tajik
Thai
Tigrinya
Tonga
Turkish
Turkmen
Uighur
Ukrainian
Uzbek
Vietnamese
Welsh
Wolof