1
00:00:00,917 --> 00:00:03,825
So how are we going to beat
this novel coronavirus?

2
00:00:04,317 --> 00:00:06,948
By using our best tools:

3
00:00:06,972 --> 00:00:09,011
our science and our technology.

4
00:00:09,594 --> 00:00:12,726
In my lab, we're using
the tools of artificial intelligence

5
00:00:12,750 --> 00:00:14,329
and synthetic biology

6
00:00:14,353 --> 00:00:17,413
to speed up the fight
against this pandemic.

7
00:00:18,078 --> 00:00:19,941
Our work was originally designed

8
00:00:19,965 --> 00:00:22,818
to tackle the antibiotic
resistance crisis.

9
00:00:22,842 --> 00:00:27,531
Our project seeks to harness
the power of machine learning

10
00:00:27,555 --> 00:00:29,401
to replenish our antibiotic arsenal

11
00:00:29,425 --> 00:00:33,263
and avoid a globally devastating
postantibiotic era.

12
00:00:33,685 --> 00:00:36,505
Importantly, the same
technology can be used

13
00:00:36,529 --> 00:00:38,601
to search for antiviral compounds

14
00:00:38,625 --> 00:00:41,303
that could help us fight
the current pandemic.

15
00:00:42,080 --> 00:00:45,982
Machine learning is turning
the traditional model of drug discovery

16
00:00:46,006 --> 00:00:47,410
on its head.

17
00:00:47,434 --> 00:00:48,659
With this approach,

18
00:00:48,683 --> 00:00:52,761
instead of painstakingly testing
thousands of existing molecules

19
00:00:52,785 --> 00:00:54,221
one by one in a lab

20
00:00:54,245 --> 00:00:55,832
for their effectiveness,

21
00:00:55,856 --> 00:01:00,513
we can train a computer
to explore the exponentially larger space

22
00:01:00,537 --> 00:01:04,121
of essentially all possible molecules
that could be synthesized,

23
00:01:04,145 --> 00:01:09,759
and thus, instead of looking
for a needle in a haystack,

24
00:01:09,783 --> 00:01:13,543
we can use the giant magnet
of computing power

25
00:01:13,567 --> 00:01:17,482
to find many needles
in multiple haystacks simultaneously.

26
00:01:18,423 --> 00:01:20,415
We've already had some early success.

27
00:01:21,010 --> 00:01:26,475
Recently, we used machine learning
to discover new antibiotics

28
00:01:26,499 --> 00:01:29,059
that can help us fight off
the bacterial infections

29
00:01:29,083 --> 00:01:32,694
that can occur alongside
SARS-CoV-2 infections.

30
00:01:33,181 --> 00:01:37,350
Two months ago, TED's Audacious Project
approved funding for us

31
00:01:37,374 --> 00:01:39,562
to massively scale up our work

32
00:01:39,586 --> 00:01:44,214
with the goal of discovering
seven new classes of antibiotics

33
00:01:44,238 --> 00:01:47,721
against seven of the world's
deadly bacterial pathogens

34
00:01:47,745 --> 00:01:49,800
over the next seven years.

35
00:01:50,206 --> 00:01:51,939
For context:

36
00:01:51,963 --> 00:01:53,891
the number of new class of antibiotics

37
00:01:53,915 --> 00:01:57,150
that have been discovered
over the last three decades is zero.

38
00:01:58,030 --> 00:02:01,601
While the quest for new antibiotics
is for our medium-term future,

39
00:02:01,625 --> 00:02:06,277
the novel coronavirus poses
an immediate deadly threat,

40
00:02:06,301 --> 00:02:10,094
and I'm excited to share that we think
we can use the same technology

41
00:02:10,118 --> 00:02:12,927
to search for therapeutics
to fight this virus.

42
00:02:13,486 --> 00:02:15,205
So how are we going to do it?

43
00:02:15,229 --> 00:02:18,177
Well, we're creating
a compound training library

44
00:02:18,201 --> 00:02:23,743
and with collaborators applying these
molecules to SARS-CoV-2-infected cells

45
00:02:23,767 --> 00:02:27,661
to see which of them exhibit
effective activity.

46
00:02:28,175 --> 00:02:31,367
These data will be use to train
a machine learning model

47
00:02:31,391 --> 00:02:35,461
that will be applied to an in silico
library of over a billion molecules

48
00:02:35,485 --> 00:02:39,689
to search for potential
novel antiviral compounds.

49
00:02:40,324 --> 00:02:42,982
We will synthesize and test
the top predictions

50
00:02:43,006 --> 00:02:45,895
and advance the most promising
candidates into the clinic.

51
00:02:46,356 --> 00:02:48,134
Sound too good to be true?

52
00:02:48,158 --> 00:02:49,590
Well, it shouldn't.

53
00:02:49,614 --> 00:02:52,939
The Antibiotics AI Project is founded
on our proof of concept research

54
00:02:52,963 --> 00:02:56,364
that led to the discovery
of a novel broad-spectrum antibiotic

55
00:02:56,388 --> 00:02:57,573
called halicin.

56
00:02:58,443 --> 00:03:01,256
Halicin has potent antibacterial activity

57
00:03:01,280 --> 00:03:05,382
against almost all antibiotic-resistant
bacterial pathogens,

58
00:03:05,406 --> 00:03:09,047
including untreatable
panresistant infections.

59
00:03:09,862 --> 00:03:12,132
Importantly, in contrast
to current antibiotics,

60
00:03:12,156 --> 00:03:15,850
the frequency at which bacteria
develop resistance against halicin

61
00:03:15,874 --> 00:03:17,358
is remarkably low.

62
00:03:18,303 --> 00:03:23,013
We tested the ability of bacteria
to evolve resistance against halicin

63
00:03:23,037 --> 00:03:24,825
as well as Cipro in the lab.

64
00:03:25,299 --> 00:03:26,841
In the case of Cipro,

65
00:03:26,865 --> 00:03:29,690
after just one day, we saw resistance.

66
00:03:30,213 --> 00:03:31,691
In the case of halicin,

67
00:03:31,715 --> 00:03:33,830
after one day,
we didn't see any resistance.

68
00:03:34,479 --> 00:03:37,781
Amazingly, after even 30 days,

69
00:03:37,805 --> 00:03:40,406
we didn't see any resistance
against halicin.

70
00:03:41,098 --> 00:03:46,624
In this pilot project, we first tested
roughly 2,500 compounds against E. coli.

71
00:03:47,259 --> 00:03:50,039
This training set included
known antibiotics,

72
00:03:50,063 --> 00:03:51,809
such as Cipro and penicillin,

73
00:03:51,833 --> 00:03:54,105
as well as many drugs
that are not antibiotics.

74
00:03:54,984 --> 00:03:57,571
These data we used to train a model

75
00:03:57,595 --> 00:04:01,573
to learn molecular features
associated with antibacterial activity.

76
00:04:02,269 --> 00:04:04,970
We then applied this model
to a drug-repurposing library

77
00:04:04,994 --> 00:04:07,472
consisting of several thousand molecules

78
00:04:07,496 --> 00:04:10,114
and asked the model to identify molecules

79
00:04:10,138 --> 00:04:12,922
that are predicted
to have antibacterial properties

80
00:04:12,946 --> 00:04:15,419
but don't look like existing antibiotics.

81
00:04:16,427 --> 00:04:21,224
Interestingly, only one molecule
in that library fit these criteria,

82
00:04:21,248 --> 00:04:23,584
and that molecule
turned out to be halicin.

83
00:04:24,444 --> 00:04:27,532
Given that halicin does not look
like any existing antibiotic,

84
00:04:27,556 --> 00:04:31,710
it would have been impossible for a human,
including an antibiotic expert,

85
00:04:31,734 --> 00:04:33,918
to identify halicin in this manner.

86
00:04:34,574 --> 00:04:37,204
Imagine now what we could do
with this technology

87
00:04:37,228 --> 00:04:38,969
against SARS-CoV-2.

88
00:04:39,783 --> 00:04:41,148
And that's not all.

89
00:04:41,172 --> 00:04:43,992
We're also using the tools
of synthetic biology,

90
00:04:44,016 --> 00:04:46,627
tinkering with DNA
and other cellular machinery,

91
00:04:46,651 --> 00:04:50,561
to serve human purposes
like combating COVID-19,

92
00:04:50,585 --> 00:04:54,232
and of note, we are working
to develop a protective mask

93
00:04:54,256 --> 00:04:57,688
that can also serve
as a rapid diagnostic test.

94
00:04:58,192 --> 00:04:59,664
So how does that work?

95
00:04:59,688 --> 00:05:00,893
Well, we recently showed

96
00:05:00,917 --> 00:05:03,860
that you can take the cellular
machinery out of a living cell

97
00:05:03,884 --> 00:05:07,976
and freeze-dry it along with
RNA sensors onto paper

98
00:05:08,000 --> 00:05:12,916
in order to create low-cost
diagnostics for Ebola and Zika.

99
00:05:13,503 --> 00:05:18,730
The sensors are activated when
they're rehydrated by a patient sample

100
00:05:18,754 --> 00:05:21,576
that could consist of blood
or saliva, for example.

101
00:05:21,600 --> 00:05:24,861
It turns out, this technology
is not limited to paper

102
00:05:24,885 --> 00:05:27,771
and can be applied
to other materials, including cloth.

103
00:05:28,671 --> 00:05:30,613
For the COVID-19 pandemic,

104
00:05:30,637 --> 00:05:34,983
we're designing RNA sensors
to detect the virus

105
00:05:35,007 --> 00:05:38,217
and freeze-drying these
along with the needed cellular machinery

106
00:05:38,241 --> 00:05:40,948
into the fabric of a face mask,

107
00:05:40,972 --> 00:05:43,201
where the simple act of breathing,

108
00:05:43,225 --> 00:05:45,502
along with the water vapor
that comes with it,

109
00:05:45,526 --> 00:05:47,286
can activate the test.

110
00:05:47,804 --> 00:05:52,064
Thus, if a patient is infected
with SARS-CoV-2,

111
00:05:52,088 --> 00:05:54,161
the mask will produce
a fluorescent signal

112
00:05:54,185 --> 00:05:58,015
that could be detected by a simple,
inexpensive handheld device.

113
00:05:58,534 --> 00:06:03,018
In one or two hours, a patient
could thus be diagnosed

114
00:06:03,042 --> 00:06:06,014
safely, remotely and accurately.

115
00:06:06,735 --> 00:06:09,255
We're also using synthetic biology

116
00:06:09,279 --> 00:06:11,999
to design a candidate
vaccine for COVID-19.

117
00:06:13,014 --> 00:06:15,667
We are repurposing the BCG vaccine,

118
00:06:15,691 --> 00:06:18,561
which had been used against TB
for almost a century.

119
00:06:18,585 --> 00:06:20,126
It's a live attenuated vaccine,

120
00:06:20,150 --> 00:06:24,807
and we're engineering it
to express SARS-CoV-2 antigens,

121
00:06:24,831 --> 00:06:27,645
which should trigger the production
of protective antibodies

122
00:06:27,669 --> 00:06:29,304
by the immune system.

123
00:06:29,328 --> 00:06:32,062
Importantly, BCG
is massively scalable

124
00:06:32,086 --> 00:06:36,659
and has a safety profile that's among
the best of any reported vaccine.

125
00:06:37,881 --> 00:06:42,986
With the tools of synthetic biology
and artificial intelligence,

126
00:06:43,010 --> 00:06:46,358
we can win the fight
against this novel coronavirus.

127
00:06:46,844 --> 00:06:50,163
This work is in its very early stages,
but the promise is real.

128
00:06:50,798 --> 00:06:54,243
Science and technology
can give us an important advantage

129
00:06:54,267 --> 00:06:57,428
in the battle of human wits
versus the genes of superbugs,

130
00:06:57,452 --> 00:06:59,199
a battle we can win.

131
00:06:59,990 --> 00:07:01,223
Thank you.