WEBVTT - Using AI Digital Twins for Drug Testing

0:00:00.160 --> 0:00:03.760
<v Speaker 1>This is Bloomberg Business Week with Carol Masser and Jason

0:00:03.840 --> 0:00:08.160
<v Speaker 1>Kelly on Bloomberg Radio. This is a really interesting discussion

0:00:08.160 --> 0:00:10.639
<v Speaker 1>we're about to have and it actually harkens back and

0:00:10.720 --> 0:00:13.440
<v Speaker 1>thanks to Carol Masser for pointing this out on Twitter

0:00:13.520 --> 0:00:16.000
<v Speaker 1>earlier to a conversation. We actually had it in j

0:00:16.160 --> 0:00:18.800
<v Speaker 1>I t earlier this year when we were still out

0:00:18.840 --> 0:00:20.840
<v Speaker 1>about out and about in the world. It was a

0:00:20.840 --> 0:00:25.720
<v Speaker 1>really live event. Yeah, they're in Newark. We were talking

0:00:25.720 --> 0:00:29.280
<v Speaker 1>about the concept of digital twins. What is that, you ask,

0:00:29.520 --> 0:00:32.240
<v Speaker 1>Let's get into it. Charles Fisher is founder and CEO

0:00:32.400 --> 0:00:35.400
<v Speaker 1>of Unlearned AI. He joins us on the phone from

0:00:35.440 --> 0:00:38.159
<v Speaker 1>San Francisco. Dr Fisher, thank you so much for joining us.

0:00:39.479 --> 0:00:41.440
<v Speaker 1>Thank you for having me. All Right, So this is

0:00:41.479 --> 0:00:45.920
<v Speaker 1>exciting in part because we are all now many experts,

0:00:46.000 --> 0:00:49.080
<v Speaker 1>or at least we consider ourselves such on drug development.

0:00:49.120 --> 0:00:51.680
<v Speaker 1>You know, we talked with the CEO of Teva Pharmaceuticals

0:00:51.720 --> 0:00:54.200
<v Speaker 1>here on this program earlier and really got into a

0:00:54.240 --> 0:00:57.720
<v Speaker 1>lot of the issues here. How does artificial intelligence play

0:00:57.800 --> 0:01:00.320
<v Speaker 1>into this? How might it help us in an age

0:01:00.320 --> 0:01:06.200
<v Speaker 1>where we're all really interested in accelerating development? Here? Yeah, well,

0:01:06.240 --> 0:01:09.080
<v Speaker 1>I think there are a few different ways, starting all

0:01:09.120 --> 0:01:11.400
<v Speaker 1>the way from the beginning of the cycle to saying

0:01:11.600 --> 0:01:16.160
<v Speaker 1>how can we discover new compounds that might be effective therapies,

0:01:16.480 --> 0:01:18.679
<v Speaker 1>all the way to the end of the process, and

0:01:18.720 --> 0:01:21.760
<v Speaker 1>saying how can we leverage all of the data that

0:01:21.840 --> 0:01:26.280
<v Speaker 1>we have from electronic health records to make more efficient

0:01:26.400 --> 0:01:32.120
<v Speaker 1>clinical trials that can get drugs to patients faster. Well, okay,

0:01:32.400 --> 0:01:34.680
<v Speaker 1>so what's interesting is, and we're hearing more and more

0:01:34.680 --> 0:01:37.360
<v Speaker 1>about this, right, We've heard various time frames and what

0:01:37.480 --> 0:01:41.800
<v Speaker 1>it will take to develop um a new vaccine, and

0:01:42.200 --> 0:01:46.120
<v Speaker 1>the latest have been anywhere from nine months maybe the

0:01:46.160 --> 0:01:49.160
<v Speaker 1>most optimistic too, maybe a couple of years. So tell us,

0:01:49.920 --> 0:01:52.520
<v Speaker 1>you know, you shared with us some some some wisdom

0:01:52.560 --> 0:01:55.960
<v Speaker 1>about when you normally develop a new medicine, it's between

0:01:55.960 --> 0:02:00.360
<v Speaker 1>ten and fifteen years. How can digital cloning help skip

0:02:00.440 --> 0:02:07.120
<v Speaker 1>through this vaccine creation process faster and also ensure that

0:02:07.160 --> 0:02:12.440
<v Speaker 1>it's safe. Yeah, I think the biggest part of the

0:02:12.520 --> 0:02:15.280
<v Speaker 1>problem in terms of the time when it takes to

0:02:15.320 --> 0:02:17.120
<v Speaker 1>talk about how long will it take us to get

0:02:17.160 --> 0:02:19.840
<v Speaker 1>new treatments or new vaccines, whether that be for for

0:02:19.919 --> 0:02:23.280
<v Speaker 1>COVID nineteen or any disease, is the amount of time

0:02:23.320 --> 0:02:25.760
<v Speaker 1>it takes for us to tell if those are safe

0:02:25.800 --> 0:02:29.080
<v Speaker 1>and effective. So you know, one of the sort of

0:02:29.320 --> 0:02:32.200
<v Speaker 1>bad parts about drug development is that actually about nine

0:02:32.200 --> 0:02:36.120
<v Speaker 1>out of ten drugs that we try in clinical trials

0:02:36.200 --> 0:02:39.440
<v Speaker 1>end up failing. So most of the time our guesses

0:02:39.680 --> 0:02:43.040
<v Speaker 1>end up not working well. So it's what we'd like

0:02:43.080 --> 0:02:44.840
<v Speaker 1>to be able to do is to speed that up,

0:02:45.360 --> 0:02:48.640
<v Speaker 1>uh and in a variety of ways by postentially leveraging

0:02:49.240 --> 0:02:51.760
<v Speaker 1>uh So, what we call these these digital patients is

0:02:51.760 --> 0:02:57.480
<v Speaker 1>is leveraging data from electronic health records to make trials faster. Alright,

0:02:57.520 --> 0:03:01.639
<v Speaker 1>so tell us how it works. Sure, So, basically, when

0:03:01.639 --> 0:03:05.320
<v Speaker 1>a patient enrolls in the trial, we create a digital

0:03:05.360 --> 0:03:09.440
<v Speaker 1>copy of that patient that tells us what would happen

0:03:09.600 --> 0:03:12.040
<v Speaker 1>to that person if they were to receive a placebo,

0:03:12.120 --> 0:03:14.480
<v Speaker 1>so a dummy treatment, and then at the end of

0:03:14.480 --> 0:03:17.800
<v Speaker 1>the study, you give that real person the real treatment

0:03:18.520 --> 0:03:20.760
<v Speaker 1>how it affects them, and then you can compare it

0:03:20.800 --> 0:03:24.240
<v Speaker 1>to the digital twins prediction for what would have happened

0:03:24.520 --> 0:03:26.920
<v Speaker 1>if they had received the placebo, and then you can

0:03:27.080 --> 0:03:30.040
<v Speaker 1>estimate if that treatment was effective or not. And basically,

0:03:30.080 --> 0:03:33.960
<v Speaker 1>because you can do this but these predicted placebo responses,

0:03:34.280 --> 0:03:36.760
<v Speaker 1>you don't need to enroll as many subjects into a

0:03:36.760 --> 0:03:38.840
<v Speaker 1>clinical trump so you can run a trial with up

0:03:38.920 --> 0:03:43.680
<v Speaker 1>to half as many half as many subjects, So doing

0:03:43.720 --> 0:03:46.520
<v Speaker 1>this are we doing this in this? In this you know,

0:03:47.360 --> 0:03:54.840
<v Speaker 1>search um for a vaccine. I'm not aware of any

0:03:55.320 --> 0:04:00.000
<v Speaker 1>trials that are doing this in COVID nineteen right now. UM.

0:04:00.040 --> 0:04:03.040
<v Speaker 1>I think one of the difficulties there is because it's

0:04:03.080 --> 0:04:06.200
<v Speaker 1>such a new disease, we need to have a lot

0:04:06.240 --> 0:04:08.600
<v Speaker 1>of data for, you know, to say we know how

0:04:08.600 --> 0:04:11.160
<v Speaker 1>this disease will progress if you don't receive a treatment.

0:04:11.480 --> 0:04:13.960
<v Speaker 1>But since COVID nineteen is new, we really don't have

0:04:14.040 --> 0:04:16.840
<v Speaker 1>that that those data yet. But lots and lots of

0:04:16.920 --> 0:04:20.360
<v Speaker 1>data are being collected every day, So the ability to

0:04:20.360 --> 0:04:23.520
<v Speaker 1>to apply these types of technologies may that may be

0:04:23.640 --> 0:04:26.720
<v Speaker 1>something that could be done in the near future. All right,

0:04:26.839 --> 0:04:29.280
<v Speaker 1>we are all aware, and you're much more aware than

0:04:29.320 --> 0:04:33.479
<v Speaker 1>we are of the trials and tribulations as it were.

0:04:33.480 --> 0:04:35.760
<v Speaker 1>When it comes to the regulatory side of this, and

0:04:35.800 --> 0:04:38.640
<v Speaker 1>a lot of the regulatory framework exists for a reason.

0:04:38.720 --> 0:04:40.640
<v Speaker 1>We want people to be safe. We want drugs when

0:04:40.640 --> 0:04:43.280
<v Speaker 1>they get to the market to be as safe as

0:04:43.279 --> 0:04:47.159
<v Speaker 1>they can possibly be. What is the reaction, what has

0:04:47.200 --> 0:04:51.560
<v Speaker 1>been the reaction by organizations and institutions like the f

0:04:51.720 --> 0:04:56.560
<v Speaker 1>D A two plans like this mhm. The FDA is

0:04:56.600 --> 0:04:59.760
<v Speaker 1>actually doing a lot of work uh these days to

0:05:00.120 --> 0:05:04.240
<v Speaker 1>modernized clinical trials really across the board, and so in

0:05:04.279 --> 0:05:07.880
<v Speaker 1>general they're they're quite supportive of new approaches to bring

0:05:07.920 --> 0:05:10.520
<v Speaker 1>in these kinds of data to make things more efficient.

0:05:10.920 --> 0:05:13.560
<v Speaker 1>You really just have to work with them to demonstrate

0:05:13.640 --> 0:05:16.640
<v Speaker 1>with real evidence that the approach you're taking works well,

0:05:16.680 --> 0:05:20.040
<v Speaker 1>which is really how it should be anyway, right, um, So,

0:05:20.279 --> 0:05:22.360
<v Speaker 1>so I would say that they're quite open to it. Uh.

0:05:22.400 --> 0:05:25.440
<v Speaker 1>The approaches are also new, so there they have been

0:05:25.560 --> 0:05:29.800
<v Speaker 1>some I think a handful now of of drugs for

0:05:30.040 --> 0:05:33.880
<v Speaker 1>targeting different types of cancer where these types of evidence

0:05:33.880 --> 0:05:36.680
<v Speaker 1>have been have been used. Um. And then you know,

0:05:37.080 --> 0:05:39.599
<v Speaker 1>our company is applying some of these approaches in some

0:05:39.720 --> 0:05:44.440
<v Speaker 1>trials for for all disease. Now wells then talk to

0:05:44.520 --> 0:05:47.000
<v Speaker 1>us a little bit about you know, maybe it's not

0:05:47.080 --> 0:05:50.080
<v Speaker 1>something that's applicable as you said to COVID nineteen and

0:05:50.120 --> 0:05:54.560
<v Speaker 1>the hunt for a vaccine, but there are certainly some diseases,

0:05:54.560 --> 0:05:58.159
<v Speaker 1>whether it's Alzheimer's and so on, that really have plagued

0:05:58.240 --> 0:06:01.080
<v Speaker 1>us and trying to figure out some kind of cure

0:06:01.240 --> 0:06:03.960
<v Speaker 1>or some kind of you know, more significant treatment than

0:06:03.960 --> 0:06:07.000
<v Speaker 1>we currently have. Talked to us attle bit about those

0:06:07.040 --> 0:06:09.440
<v Speaker 1>types of ailments that might there might be some promise

0:06:09.560 --> 0:06:14.480
<v Speaker 1>using digital clones. Yeah, I think in you know, lots

0:06:14.520 --> 0:06:18.000
<v Speaker 1>of these diseases, whether they be you know, cancer, or

0:06:18.160 --> 0:06:22.720
<v Speaker 1>neurologic disease like Alzheimer's, or or rare genetic diseases. UM,

0:06:22.800 --> 0:06:25.400
<v Speaker 1>the ability to leverage all of the data we have

0:06:25.760 --> 0:06:28.680
<v Speaker 1>collected on those diseases to make those trials more efficient

0:06:29.279 --> 0:06:31.279
<v Speaker 1>UM with digital twins is I think going to be

0:06:31.400 --> 0:06:33.440
<v Speaker 1>very Uh. I think that's going to be the future

0:06:33.480 --> 0:06:36.160
<v Speaker 1>of how trials are run. It also, you know, one

0:06:36.200 --> 0:06:39.520
<v Speaker 1>of the sort of side effects of COVID nineteen that

0:06:39.560 --> 0:06:42.239
<v Speaker 1>since we're all sheltering in place right now, no one's

0:06:42.279 --> 0:06:45.880
<v Speaker 1>participating in clinical trial right so clinical trials for all

0:06:45.920 --> 0:06:49.080
<v Speaker 1>of these other diseases, for Alzheimers, for cancer, they're all

0:06:49.120 --> 0:06:52.840
<v Speaker 1>disruptive right now. And so figuring out ways to run

0:06:52.880 --> 0:06:57.400
<v Speaker 1>those trials with fewer patients or where so that patients

0:06:57.440 --> 0:07:00.640
<v Speaker 1>don't have to maybe go into the hospital by applying

0:07:00.680 --> 0:07:03.919
<v Speaker 1>new technologies like like digital twins will really help to

0:07:04.040 --> 0:07:07.200
<v Speaker 1>keep medical research going so that it's not all set

0:07:07.240 --> 0:07:11.160
<v Speaker 1>back because we aren't able to participate because of COVID

0:07:11.200 --> 0:07:15.200
<v Speaker 1>Night Team. All right, well, really interesting, interested to see

0:07:15.200 --> 0:07:17.640
<v Speaker 1>where this goes from. Here are thanks to Charles Fisher,

0:07:17.920 --> 0:07:20.960
<v Speaker 1>founders CEO of Unlearned AI, joining us on the phone

0:07:21.280 --> 0:07:24.600
<v Speaker 1>from San Francisco and the idea of digital twins. It

0:07:24.720 --> 0:07:28.360
<v Speaker 1>is cool, very cool. It makes sense, yes, and we

0:07:28.440 --> 0:07:29.840
<v Speaker 1>need to do more on it because I do think

0:07:29.840 --> 0:07:31.480
<v Speaker 1>this is going to be a big way for the

0:07:31.560 --> 0:07:34.480
<v Speaker 1>medical arena and their way forward to in terms of

0:07:34.520 --> 0:07:35.560
<v Speaker 1>tackling a lot of ailments.