How Pharma Is Using AI For Clinical Drug Development

Featured On YouTube - Best of the Web

Filming Date: May 2020


*AI Generated

So thank you for the introduction. And I’ll be talking about applying AI in early clinical development of new drugs. So if you work in Pharma or clinical trials, clinical development development, you probably know many of these types of big data. But this is actually not true. Ten years ago, if you think about it, genome sequencing was already available as a technique, but it was still too expensive and not convenient enough to use in clinical trials.

But today, actually, we are routinely use genome sequencing of patients in clinical trials to get further insight, to find the biomarkers. Single cell sequencing was not even available at that time. And if you look at all other data types, they were either not there at that time or the technology was there but not really widely used. So that includes, for example, realworld data such as EHRs, insurance claims, digital imaging wearables or sensors and clinical biomarkers. But today all of these are available and they are actually being powered by deep learning or other AI techniques.

So we actually in clinical development can do a lot more than we used to be able to. So I want to focus on the early clinical development of new drugs. And this is where the translational research happens. And by early, I mean the phase one and the phase two clinical trials. This is where you assess the safety and early efficacy proof of concept, whether the drug is really going to work before scaling that up in phase three.

So it’s a very unique phase where you actually test the hypothesis that you develop in the preclinical stage for the first time in the real patients. So the success or failure of this translation really has an outsized impact on the overall success or failure of your drug. And you can also imagine any technology that can improve the success rate will have really large impact on drug development. So the natural question is, can AI improve this success rate of drug development? And we believe I think you probably all do.

The answer is yes. And there are two ways this can happen. So one is impacting the science side of translational research. The other is the operational side. By science, it’s mainly impacting finding better biomarkers to develop more precise medicines that’s really about is the precision medicine really what translational research is about?

But we don’t want to forget the operational side where the clinical trials can actually be hugely improved. As you know, clinical trial is very expensive and very efficient. Anything we can do to make that more efficient, make the data more reliable, that can improve the translational research tremendously.

So we have been doing AI for the past several years and where are we in this journey? And I would like to argue for Rose and maybe for other Pharma as well. We have passed definitely the first stage demonstrating the value and understanding limitations, because in the early stage we really wanted to know, is AI really going to be able to deliver as people say? And there are a number of examples I can show there. And the second is once we demonstrate the value, can we explore further opportunities?

Can we find further potential to really enlarge this impact? And I think someone already alluded to earlier that culture is really a barrier. But because in pharmaceutical company, we value molecules. That’s what we patent, that’s what we develop into clinical stage and market. And really eventually that’s what helped patients.

But I think the culture is slowly changing that we want to value data as much as a molecule because this data can generate insights, can make our molecule better, can make our clinical trials more efficient, and can do a lot of innovative things that we haven’t thought about before. And the last stage, I think, is we want to operationalise AI, meaning here. I mean really embed AI in the whole operations, the science operations of Pharma research and development. And we are not there yet.

So for the rest of the talk, I’m going to just go through this and then provide some examples for each of the categories. And in my first example, so I want to talk about using deep learning and neurological diseases. So this is one of our early success stories in demonstrating the power of deep learning. This happens I think maybe four years ago started and as you know, Parkinson’s disease, this is a disease really affects celebrities but can affect any of us sitting here. Unfortunately, the way to diagnose and to monitor patient symptoms if you are in a clinical trial, it’s a very old and crude method.

This is up thers testing. Basically, you go to the doctor’s office. There are a series of tests, behavioral tests or movement tests that you go through. So there are two problem with this. One is it could be very subjective because a different physician may come up with different answers, could be affected by the patient move that day or something else that day.

And second is that we don’t really know what happens between each visit to the clinic. So in the bottom right, I think panel here is representation of patient journey in the whole year. So 365 bars there each represent one day. And you can see here are the two visits. Basically in this, this is a phase one trial that we run for about a year and that you have two visits.

And the problem is that the symptoms really the red dots represent the severity of the symptoms of the patient over a year. But the patient can really only recall about maybe a few days, a week of what happened between the two visits. So when the doctor asks how was your symptoms since the last visit, you don’t really remember what happened. Two weeks, three weeks, two months, three months back, you’re missing a lot of information. And second, as I mentioned, it’s really not that accurate because there is no molecular biomarker that you can use.

So if you look at typical curve, actually, this is maybe ideal situation that you could see difference between placebo and the drug. But in reality, you may see a very large arrow bar, and you actually also don’t know what’s happening in between. So that doesn’t help us in understanding how the drug is really doing.

So the question is, can we use any digital technology here? We call a digital biomarker, basically using the sensor signal from the smartphone to answer that kind of question. So we devised two type of tests. One is called active test, which mimics the UPDRS testing. And so, for example, the patient will walk, we’ll do a balance, we’ll do rest and tremor.

Those things are being measured. And the second part of the test is passive monitoring, where patients just put the cell phone in their pockets and wear that with them during the day, and we stream the data and collect the data and then see what we can learn from that. So this is what the data would look like. The cell phone actually generates a lot of data continuously, and that’s both generation and acceleration in three dimensions. And this is actually one of our colleagues in Basel, Switzerland.

He didn’t know he was going to be filmed on that day when he was wearing shorts. And this is actually the deep learning model our colleague we in our New York site developed. So he took a deep learning model from the public, and he actually adapted that to the sensor data in our clinical trials. And he used a 50 hours of public sensor data to train his model.

And in the validation, the held out test, he achieved very high accuracy, 99%. 98% accuracy. And when this is applied to our patients. So we were using the active test, as the validation says, because in the active test, we actually know whether the patient is sitting.

So actually, between standing and walking, we were able to achieve 99% or 96% accuracy. And in reality, we also are predicting multi classes. It’s patient lining, standing, sitting, it’s walking stairs, et cetera.

This is some preliminary results in the phase one trial. So as you see that indeed, you could see the difference purely based on the phone data, not based on any physician assessment, to tell the difference between healthy volunteers and Parkinson’s disease patients. So both, you know that patients spend less time walking, and when they do stand, they actually have a less stand, and they also invest less power in their terms. So these are things that can you actually gain from your model. And the patients you could see that also patients, they walk less, they invest less power in the walking, and they walk slower.

So these are things that were in our first example. But we were now in the phase to expand that in many different neurological diseases. One example is schizophrenia. He used the same model trained on different data. So this is actually graphic watch with the watch data and then similarly achieve the very high accuracy.

This is the correlation between predicted and actual results. And to make a long story short, you could also see correlations between the activity ratio. And basically this is a measure of how motivated the patients are in their life and also some negative symptoms such as diminished expression. So the patients who are more active in their gesture, the power of their gesture, they actually are doing better in terms of their clinical symptoms. And this, as you can imagine, can eventually become a more objective way of measuring patients in clinic and in many ways in helping our drug development.

So moving on to another disease area and different data types. So you saw the digital biomarkers. We could also ask, can AI be used in traditional molecular biomarkers? For example, in oncology and oncology, we have abundance of molecular biomarkers candidates, usually hundreds if not thousands in clinical trial, phase one, phase two. But the validated markers are still very rare and especially in immunotherapy where you don’t know the mechanism of action as well.

So this is Francesca in our New York team also did she actually took all the baseline lab readouts that includes both the immune biomarkers and also the chemistry biomarkers and using the gradient boosting model and then try to see which of the features are possible biomarkers that can predict patient response, for example, complete response, partial response, stable disease or progressive disease. And here, since we actually didn’t have that many patients, we couldn’t really build a predictive model. But we use the machine learning approach to rank the features based on their importance. And here you can see there are just two examples of the features. One marker is actually immune marker, the other is actually a chemistry marker.

Actually, each one of them have some predictive power, but by combining them, actually you could see that these are the ones that patients that are responding and these are the patients that don’t respond. So by having two biomarkers, you actually can have much better prediction. You could also ask, are they really just some random features that you select? So we actually use our real world data, the electronic health record from Flatiron that we use to see can we validate some of these? So this is a result of one of the lab chemistry markers that we could actually see that patients with lower this chemistry marker activity have higher survival chance.

And this we actually find in quite a few of the markers that we found. Basically, patients who are healthier in various ways, they actually do better. Not only they live longer, they actually respond to the drug better. And these are actually could be very useful in predicting which patients might respond to certain drugs. And I mentioned operation size of the translational research I won’t go into too much detail here but I think the key point here is that if we look at one example, patient recruitment this is really the most expensive part of running clinical trials more than 30% of total cost but it’s very efficient because less than 10% of patients actually you see really complete clinical trials there are multiple points where patients are either disqualified, drop out or just leave the clinical trials So as you can imagine, if we can improve the patient recruitment, have the right patients at the right clinical sites then our data can become a lot more reliable, our science can become also much better and we actually explore the real world data to improve patient recruitment that actually worked and my time is up and we actually also looking at can we use AI to actually help us do more predict?

Much better prediction? So I’m going through the next few slides very quickly I mentioned culture here as you know that in Pharma we want to introduce this data culture and one of the things that was done was introducing the data challenges or hackathons that are very prevalent in machine learning community but not so much of a practice in Pharma and this is the first Red challenge or the advanced analytics data challenge we use flat earning data with the goal of predicting patient survival after one year of treatment and I just want to point to really 500 people from 28 roadsides attended this with 132 teams really great participation and this is another example that we actually our HR is also thinking about how to recruit the new kinds of challenges so they put the code for life challenge for the community and also this type of thing that could be used in recruitment of talent this is, I think my loss and basically I think we are still not there in operationalizing AI and there are two major things one is we still like data, we still like well labeled, well curated data and second is we don’t even have the infrastructure to really make data fair so that we mean, can data be findable, be accessible, interoperable or reusable?

Without that, it’s actually very difficult to do large scale AI so that’s my end and I think I have demonstrated tremendous potential and also we are exploring different opportunities and I hope that new culture is taking shape and there’s a lot of work that we still need to do in operationalizing AI and I will stop here and we’ll you are thinking of questions so I just want to acknowledge quite a few people there.


Subscribe to our newsletter

Sign up to receive updates, promotions, and sneak peaks of upcoming event. Plus 20% off your next order.

Promotion nulla vitae elit libero a pharetra augue