The medical potential of AI and metabolites | Leila Pirhaji

68,961 views ใƒป 2019-11-20

TED


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืชืจื’ื•ื: zeeva livshitz ืขืจื™ื›ื”: Ido Dekkers
ื‘ 2003,
ื›ืืฉืจ ืจื™ืฆืคื ื• ืืช ื”ื’ื ื•ื ื”ืื ื•ืฉื™,
ื—ืฉื‘ื ื• ืฉืชื™ืžืฆื ืœื ื• ื”ืชืฉื•ื‘ื” ืœื˜ื™ืคื•ืœ ื‘ืžื—ืœื•ืช ืจื‘ื•ืช.
00:13
In 2003,
0
13507
1889
ืื‘ืœ ื”ืžืฆื™ืื•ืช ืจื—ื•ืงื” ืžื›ืš,
00:15
when we sequenced the human genome,
1
15420
2913
00:18
we thought we would have the answer to treat many diseases.
2
18357
3922
ื›ื™ ื‘ื ื•ืกืฃ ืœื’ื ื™ื ืฉืœื ื•,
ืœืกื‘ื™ื‘ื” ื•ืœืื•ืจื— ื”ื—ื™ื™ื ืฉืœื ื• ื™ื›ื•ืœื™ื ืœื”ื™ื•ืช ืชืคืงื™ื“ ืžืฉืžืขื•ืชื™
00:22
But the reality is far from that,
3
22974
2657
ื‘ืคื™ืชื•ื— ืžื—ืœื•ืช ื—ืžื•ืจื•ืช ืจื‘ื•ืช.
00:26
because in addition to our genes,
4
26782
1921
ื“ื•ื’ืžื” ืื—ืช ื”ื™ื ืžื—ืœืช ื”ื›ื‘ื“ ื”ืฉื•ืžื ื™,
00:28
our environment and lifestyle could have a significant role
5
28727
4570
ืฉืžืฉืคื™ืขื” ืขืœ ืœืžืขืœื” ืž-20 ืื—ื•ื– ืžืื•ื›ืœื•ืกื™ื™ืช ื”ืขื•ืœื,
00:33
in developing many major diseases.
6
33321
2548
00:35
One example is fatty liver disease,
7
35893
3580
ื•ืื™ืŸ ืœื” ื˜ื™ืคื•ืœ ื•ื”ื™ื ืžื•ื‘ื™ืœื” ืœืกืจื˜ืŸ ื”ื›ื‘ื“
ืื• ืœืื™ ืกืคื™ืงืช ื›ื‘ื“.
00:39
which is affecting over 20 percent of the population globally,
8
39497
4083
ืื– ืจื™ืฆื•ืฃ DNA ืœื‘ื“ ืœื ื ื•ืชืŸ ืœื ื• ืžืกืคื™ืง ืžื™ื“ืข
00:43
and it has no treatment and leads to liver cancer
9
43604
3034
ื›ื“ื™ ืœืžืฆื•ื ืชืจื•ืคื•ืช ืืคืงื˜ื™ื‘ื™ื•ืช.
00:46
or liver failure.
10
46662
1423
ื‘ืฆื“ ื”ื—ื™ื•ื‘ื™, ื™ืฉื ืŸ ืžื•ืœืงื•ืœื•ืช ืจื‘ื•ืช ืื—ืจื•ืช ื‘ื’ื•ืคื ื•.
00:49
So sequencing DNA alone doesn't give us enough information
11
49517
4744
ืœืžืขืฉื”, ื™ืฉ ืžืขืœ 100,000 ืžื˜ื‘ื•ืœื™ื˜ื™ื.
00:54
to find effective therapeutics.
12
54285
2232
ืžื˜ื‘ื•ืœื™ื˜ื™ื ื”ื ื›ืœ ื”ืžื•ืœืงื•ืœื•ืช ื”ืกื•ืคืจ ืงื˜ื ื•ืช ื‘ื’ื•ื“ืœืŸ.
00:56
On the bright side, there are many other molecules in our body.
13
56541
3756
01:00
In fact, there are over 100,000 metabolites.
14
60321
3980
ื“ื•ื’ืžืื•ืช ื™ื“ื•ืขื•ืช ื”ืŸ ื’ืœื•ืงื•ื–, ืคืจื•ืงื˜ื•ื–, ืฉื•ืžื ื™ื, ื›ื•ืœืกื˜ืจื•ืœ --
01:04
Metabolites are any molecule that is supersmall in their size.
15
64325
4296
ื“ื‘ืจื™ื ืฉืื ื—ื ื• ืฉื•ืžืขื™ื ื›ืœ ื”ื–ืžืŸ.
ืžื˜ื‘ื•ืœื™ื˜ื™ื ืžืขื•ืจื‘ื™ื ื‘ื—ื™ืœื•ืฃ ื”ื—ื•ืžืจื™ื ืฉืœื ื•.
01:09
Known examples are glucose, fructose, fats, cholesterol --
16
69193
4972
ื”ื ื’ื ื‘ืžื•ืจื“ ื”ื–ืจื ืฉืœ ื”-DNA,
01:14
things we hear all the time.
17
74189
1510
ื›ืš ืฉื”ื ื ื•ืฉืื™ื ืžื™ื“ืข ื”ืŸ ืžื”ื’ื ื™ื ืฉืœื ื• ื•ื”ืŸ ืžืื•ืจื— ื”ื—ื™ื™ื.
01:16
Metabolites are involved in our metabolism.
18
76273
2983
01:20
They are also downstream of DNA,
19
80066
4028
ื”ื‘ื ืช ืžื˜ื‘ื•ืœื™ื˜ื™ื ื—ื™ื•ื ื™ืช ืœืžืฆื™ืืช ืชืจื•ืคื•ืช ืœืžื—ืœื•ืช ืจื‘ื•ืช.
01:24
so they carry information from both our genes as well as lifestyle.
20
84118
5082
ืชืžื™ื“ ืจืฆื™ืชื™ ืœื˜ืคืœ ื‘ื—ื•ืœื™ื.
01:29
Understanding metabolites is essential to find treatments for many diseases.
21
89224
5649
ืœืžืจื•ืช ื–ืืช, ืœืคื ื™ 15 ืฉื ื”, ืขื–ื‘ืชื™ ืืช ื‘ื™ืช ื”ืกืคืจ ืœืจืคื•ืื”,
ื›ื™ ื”ืชื’ืขื’ืขืชื™ ืœืžืชืžื˜ื™ืงื”.
01:34
I've always wanted to treat patients.
22
94897
2212
ื–ืžืŸ ืงืฆืจ ืœืื—ืจ ืžื›ืŸ, ืžืฆืืชื™ ืืช ื”ื“ื‘ืจ ื”ื›ื™ ืžื’ื ื™ื‘:
01:37
Despite that, 15 years ago, I left medical school,
23
97934
3858
ืื ื™ ื™ื›ื•ืœื” ืœื”ืฉืชืžืฉ ื‘ืžืชืžื˜ื™ืงื” ื›ื“ื™ ืœืœืžื•ื“ ืจืคื•ืื”.
01:41
as I missed mathematics.
24
101816
1965
ืžืื–, ืื ื™ ืžืคืชื—ืช ืืœื’ื•ืจื™ืชืžื™ื ืœื ื™ืชื•ื— ื ืชื•ื ื™ื ื‘ื™ื•ืœื•ื’ื™ื™ื.
01:45
Soon after, I found the coolest thing:
25
105019
2936
01:48
I can use mathematics to study medicine.
26
108692
2763
ืื– ื–ื” ื ืฉืžืข ืงืœ:
ื‘ื•ืื• ื•ื ืืกื•ืฃ ื ืชื•ื ื™ื ืžื›ืœ ื”ืžื˜ื‘ื•ืœื™ื˜ื™ื ื‘ื’ื•ืคื ื•,
01:53
Since then, I've been developing algorithms to analyze biological data.
27
113026
5213
ื ืคืชื— ืžื•ื“ืœื™ื ืžืชืžื˜ื™ื™ื ื›ื“ื™ ืœืชืืจ ื›ื™ืฆื“ ื”ื ืžืฉืชื ื™ื ื‘ืžื—ืœื”
01:59
So, it sounded easy:
28
119092
2283
02:01
let's collect data from all the metabolites in our body,
29
121399
3601
ื•ืœื”ืชืขืจื‘ ื‘ืฉื™ื ื•ื™ื™ื ืืœื” ืœืฉื ื˜ื™ืคื•ืœ ื‘ื”ื.
02:05
develop mathematical models to describe how they are changed in a disease
30
125024
5128
ื•ืื– ื”ื‘ื ืชื™ ืœืžื” ืืฃ ืื—ื“ ืœื ืขืฉื” ืืช ื–ื” ื‘ืขื‘ืจ:
02:10
and intervene in those changes to treat them.
31
130176
2988
ื–ื” ืงืฉื” ื‘ื™ื•ืชืจ.
(ืฆื—ื•ืง)
ื™ืฉื ื ืžื˜ื‘ื•ืœื™ื˜ื™ื ืจื‘ื™ื ื‘ื’ื•ืคื ื•.
02:14
Then I realized why no one has done this before:
32
134488
3472
ื›ืœ ืื—ื“ ืฉื•ื ื” ืžื”ืื—ืจ.
ื‘ืžื˜ื‘ื•ืœื™ื˜ื™ื ืžืกื•ื™ืžื™ื, ืื ื—ื ื• ื™ื›ื•ืœื™ื ืœืžื“ื•ื“ ืืช ื”ืžืกื” ื”ืžื•ืœืงื•ืœืจื™ืช ืฉืœื”ื
02:19
it's extremely difficult.
33
139230
1687
02:20
(Laughter)
34
140941
1087
02:22
There are many metabolites in our body.
35
142052
2412
ื‘ืืžืฆืขื•ืช ืžื›ืฉื™ืจื™ ืกืคืงื˜ืจื•ืžื˜ืจื™ื™ืช ืžืกื”.
02:24
Each one is different from the other one.
36
144783
2500
ืื‘ืœ ื‘ื’ืœืœ ืฉื™ื›ื•ืœื•ืช ืœื”ื™ื•ืช, ื› 10 ืžื•ืœืงื•ืœื•ืช ืขื ืื•ืชื” ืžืกื” ื‘ื“ื™ื•ืง,
02:27
For some metabolites, we can measure their molecular mass
37
147307
3728
ืื ื—ื ื• ืœื ื™ื•ื“ืขื™ื ื‘ื“ื™ื•ืง ืžื” ื”ืŸ,
02:31
using mass spectrometry instruments.
38
151059
2593
ื•ืื ืจื•ืฆื™ื ืœื–ื”ื•ืช ืืช ื›ื•ืœื ื‘ื‘ื™ืจื•ืจ,
02:33
But because there could be, like, 10 molecules with the exact same mass,
39
153676
4393
ืฆืจื™ืš ืœืขืฉื•ืช ืขื•ื“ ื ื™ืกื•ื™ื™ื, ืžื” ืฉื™ื›ื•ืœ ืœืงื—ืช ืขืฉืจื•ืช ืฉื ื™ื
ื•ืžื™ืœื™ืืจื“ื™ ื“ื•ืœืจื™ื.
02:38
we don't know exactly what they are,
40
158093
1807
02:39
and if you want to clearly identify all of them,
41
159924
2774
ืื– ืคื™ืชื—ื ื• ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช, ืื• ืคืœื˜ืคื•ืจืžืช ื‘โ€œืž, ื›ื“ื™ ืœืขืฉื•ืช ื–ืืช.
02:42
you have to do more experiments, which could take decades
42
162722
3104
02:45
and billions of dollars.
43
165850
1714
ืžื™ื ืคื ื• ืืช ื”ืฆืžื™ื—ื” ืฉืœ ื ืชื•ื ื™ื ื‘ื™ื•ืœื•ื’ื™ื™ื
02:48
So we developed an artificial intelligence, or AI, platform, to do that.
44
168207
5563
ื•ื‘ื ื™ื ื• ืžืกื“ ื ืชื•ื ื™ื ืฉืœ ื›ืœ ื”ืžื™ื“ืข ื”ืงื™ื™ื ืขืœ ืžื˜ื‘ื•ืœื™ื˜ื™ื
ื•ื”ืื™ื ื˜ืจืืงืฆื™ื•ืช ืฉืœื”ื ืขื ืžื•ืœืงื•ืœื•ืช ืื—ืจื•ืช.
02:53
We leveraged the growth of biological data
45
173794
2844
ืฉื™ืœื‘ื ื• ืืช ื›ืœ ื”ื ืชื•ื ื™ื ื”ืืœื” ื›ืžื’ื”-ืจืฉืช.
02:56
and built a database of any existing information about metabolites
46
176662
4424
ื•ืื–, ืžืจืงืžื•ืช ืื• ื“ื ืฉืœ ื—ื•ืœื™ื,
03:01
and their interactions with other molecules.
47
181110
3128
ืื ื• ืžื•ื“ื“ื™ื ืžืกื•ืช ืฉืœ ืžื˜ื‘ื•ืœื™ื˜ื™ื
03:04
We combined all this data as a meganetwork.
48
184262
3424
ื•ืžื•ืฆืื™ื ืืช ื”ืžืกื•ืช ืฉืžืฉืชื ื•ืช ื‘ืžื—ืœื•ืช
03:07
Then, from tissues or blood of patients,
49
187710
3396
ืื‘ืœ, ื›ืคื™ ืฉืฆื™ื™ื ืชื™ ืงื•ื“ื, ืื ื—ื ื• ืœื ื™ื•ื“ืขื™ื ื‘ื“ื™ื•ืง ืžื” ื”ืŸ.
03:11
we measure masses of metabolites
50
191130
2751
ืžืกื” ืžื•ืœืงื•ืœืจื™ืช ืฉืœ 180 ื™ื›ื•ืœื” ืœื”ื™ื•ืช ื’ืœื•ืงื•ื–, ื’ืœืงื˜ื•ื– ืื• ืคืจื•ืงื˜ื•ื–.
03:13
and find the masses that are changed in a disease.
51
193905
3259
03:17
But, as I mentioned earlier, we don't know exactly what they are.
52
197188
3190
ืœื›ื•ืœื ื™ืฉ ืืช ืื•ืชื” ืžืกื” ื‘ื“ื™ื•ืง
ืื‘ืœ ืชืคืงื•ื“ื™ื ืฉื•ื ื™ื ื‘ื’ื•ืคื ื•.
03:20
A molecular mass of 180 could be either the glucose, galactose or fructose.
53
200402
5135
ืืœื’ื•ืจื™ืชื ื”-AI ืฉืœื ื• ืฉืงืœ ืืช ื›ืœ ื”ื“ื• ืžืฉืžืขื•ื™ื•ืช ื”ืœืœื•.
ื•ืœืื—ืจ ืžื›ืŸ ื”ื•ื ื›ืจื” ืืช ื”ืžื’ื”-ืจืฉืช ื”ื–ื•
03:25
They all have the exact same mass
54
205561
2019
03:27
but different functions in our body.
55
207604
2087
ื›ื“ื™ ืœืžืฆื•ื ืื™ืš ื”ืžืกื•ืช ื”ืžื˜ื‘ื•ืœื™ื•ืช ื”ืืœื• ืžื—ื•ื‘ืจื•ืช ื–ื• ืœื–ื•
03:29
Our AI algorithm considered all these ambiguities.
56
209715
3587
ืžื” ืฉื’ื•ืจื ืœืžื—ืœื”.
03:33
It then mined that meganetwork
57
213326
2736
ื•ื‘ืขื–ืจืช ื”ืื•ืคืŸ ืฉื‘ื• ื”ืŸ ืžื—ื•ื‘ืจื•ืช,
03:36
to find how those metabolic masses are connected to each other
58
216086
4353
ื ื•ื›ืœ ืœื”ืกื™ืง ืžื”ื™ ื›ืœ ืžืกืช ืžื˜ื‘ื•ืœื™ื˜,
03:40
that result in disease.
59
220463
1958
ื›ืžื• ืœืžืฉืœ ืฉ 180 ื›ืืŸ ื™ื›ื•ืœื•ืช ืœื”ื™ื•ืช ื’ืœื•ืงื•ื–,
03:42
And because of the way they are connected,
60
222445
2238
ื•ื—ืฉื•ื‘ ืžื›ืš, ืœื’ืœื•ืช
03:44
then we are able to infer what each metabolite mass is,
61
224707
4323
ื›ื™ืฆื“ ืฉื™ื ื•ื™ื™ื ื‘ื’ืœื•ืงื•ื– ื•ื‘ืžื˜ื‘ื•ืœื™ื˜ื™ื ืื—ืจื™ื
03:49
like that 180 could be glucose here,
62
229054
2924
ืžื•ืœื™ื›ื™ื ืœืžื—ืœื”.
ื”ื”ื‘ื ื” ื”ื—ื“ืฉื ื™ืช ื”ื–ื• ืฉืœ ืžื ื’ื ื•ื ื™ ืžื—ืœื”
03:52
and, more importantly, to discover
63
232002
2551
03:54
how changes in glucose and other metabolites
64
234577
3367
ืžืืคืฉืจืช ืœื ื• ืœื’ืœื•ืช ื˜ื™ืคื•ืœื™ื ื™ืขื™ืœื™ื ื›ื“ื™ ืœื”ืชืžืงื“ ื‘ื”ื.
03:57
lead to a disease.
65
237968
1505
03:59
This novel understanding of disease mechanisms
66
239497
2995
ืื– ื”ืงืžื ื• ื—ื‘ืจืช ืกื˜ืืจื˜-ืืค ื›ื“ื™ ืœื”ื‘ื™ื ืืช ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ื”ื–ื• ืœืฉื•ืง
04:02
then enable us to discover effective therapeutics to target that.
67
242516
4492
ื•ืœื”ืฉืคื™ืข ืขืœ ื—ื™ื™ื”ื ืฉืœ ืื ืฉื™ื.
ื›ืขืช ื”ืฆื•ื•ืช ืฉืœื™ ื•ืื ื™ ื‘- ReviveMed ืขื•ื‘ื“ื™ื ืœื’ืœื•ืช
04:07
So we formed a start-up company to bring this technology to the market
68
247601
3845
ืชืจื•ืคื•ืช ืœื˜ื™ืคื•ืœ ื‘ืžื—ืœื•ืช ืงืฉื•ืช ืฉืขื‘ื•ืจืŸ ืžื˜ื‘ื•ืœื™ื˜ื™ื ื”ื ื’ื•ืจืžื™ ืžืคืชื—,
04:11
and impact people's lives.
69
251470
1805
04:13
Now my team and I at ReviveMed are working to discover
70
253722
3545
ื›ืžื• ืžื—ืœืช ื›ื‘ื“ ืฉื•ืžื ื™,
ื›ื™ ื”ื™ื ื ื’ืจืžืช ืขืœ ื™ื“ื™ ื”ืฆื˜ื‘ืจื•ืช ืฉื•ืžื ื™ื,
04:17
therapeutics for major diseases that metabolites are key drivers for,
71
257291
5105
ืฉื”ื ืกื•ื’ื™ื ืฉืœ ืžื˜ื‘ื•ืœื™ื˜ื™ื ื‘ื›ื‘ื“.
ื›ืคื™ ืฉืฆื™ื™ื ืชื™ ืงื•ื“ื, ื–ื• ืžื’ื™ืคื” ืขื ืงื™ืช ืฉืื™ืŸ ืœื” ื˜ื™ืคื•ืœ.
04:22
like fatty liver disease,
72
262420
1897
04:24
because it is caused by accumulation of fats,
73
264341
2924
ื•ืžื—ืœืช ื›ื‘ื“ ืฉื•ืžื ื™ ื”ื™ื ืจืง ื“ื•ื’ืžื” ืื—ืช.
04:27
which are types of metabolites in the liver.
74
267289
2473
ื›ื“ื™ ืœื”ืชืงื“ื, ืื ื—ื ื• ืขื•ืžื“ื™ื ืœื”ืชืžื•ื“ื“ ืขื ืžืื•ืช ืžื—ืœื•ืช ืื—ืจื•ืช
04:29
As I mentioned earlier, it's a huge epidemic with no treatment.
75
269786
3940
ืฉืื™ืŸ ืœื”ืŸ ื˜ื™ืคื•ืœ.
04:33
And fatty liver disease is just one example.
76
273750
2724
ื•ืขืœ ื™ื“ื™ ืื™ืกื•ืฃ ืขื•ื“ ื•ืขื•ื“ ื ืชื•ื ื™ื ืขืœ ืžื˜ื‘ื•ืœื™ื˜ื™ื
04:36
Moving forward, we are going to tackle hundreds of other diseases
77
276498
4178
ื•ื”ื‘ื ืช ื”ืื•ืคืŸ ื‘ื• ืฉื™ื ื•ื™ื™ื ื‘ืžื˜ื‘ื•ืœื™ื˜ื™ื
04:40
with no treatment.
78
280700
1493
ืžื•ื‘ื™ืœื™ื ืœื”ืชืคืชื—ื•ืช ืžื—ืœื•ืช,
04:42
And by collecting more and more data about metabolites
79
282217
4554
ื”ืืœื’ื•ืจื™ืชืžื™ื ืฉืœื ื• ื™ื™ืขืฉื• ื™ื•ืชืจ ื•ื™ื•ืชืจ ื—ื›ืžื™ื
04:46
and understanding how changes in metabolites
80
286795
3544
ื›ื“ื™ ืœื’ืœื•ืช ืืช ื”ืชืจื•ืคื•ืช ื”ื ื›ื•ื ื•ืช ืœืžื˜ื•ืคืœื™ื ื”ื ื›ื•ื ื™ื.
04:50
leads to developing diseases,
81
290363
2402
04:52
our algorithms will get smarter and smarter
82
292789
3489
ื•ืื ื—ื ื• ื ืชืงืจื‘ ื™ื•ืชืจ ืœื”ื’ื™ืข ืœื—ื–ื•ืŸ ืฉืœื ื•
ืฉืœ ื”ืฆืœืช ื—ื™ื™ื ืขื ื›ืœ ืฉื•ืจื” ืฉืœ ืงื•ื“.
04:56
to discover the right therapeutics for the right patients.
83
296302
4196
ืชื•ื“ื”.
05:00
And we will get closer to reach our vision
84
300522
3770
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
05:04
of saving lives with every line of code.
85
304316
3863
05:08
Thank you.
86
308203
1321
05:09
(Applause)
87
309548
3827
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7