This episode explores how artificial intelligence and machine learning allow biopharmaceutical companies to quickly and cost-effectively bring new products to market. This exciting new technology unlocks the door for discovering therapies in rare diseases and other unmet medical conditions that were once too costly and time consuming to develop.
Dr. Solomon: As opposed to high throughput screening and other ways that are more laborious on one hand and non-obvious on the other that AI might take us to the promised land if you will better, faster, cheaper simply by doing predictive work. It is like in the game Goer and like in chess, AI allows machines to win.
Jeffrey Freedman: Artificial intelligence or AI and machine learning are two buzz words that sound like they are out of a science fiction movie. But in fact this advanced technology is reshaping the way biotech startups and big pharmaceutical companies are conducting their drug discovery and drug development programs. 95% of rare diseases don’t have a single FDA treatment. The reason for this is traditionally pharmaceutical companies don’t focus their efforts on treatments for these types of diseases because the return on investment or ROI, it doesn’t warrant the time and the cost it takes to develop these therapies. It’s just too expensive. So the need for a technology that can lower the cost for drug development and shorten the time period to bring a therapy through clinical trials is huge.
Jeffrey: Welcome to the RP HealthCast. Science, Innovation, Life. One story at a time. And I am your host, Jeffrey Freedman.
Jeffrey: First to get a better understanding of AI, I spoke with Ed Miseta the executive editor of Clinical Leader and Life Science Leader magazine. Ed, can you start us out here? Let’s start with the basics. What is AI and how does it work?
Ed: So if you were to go online and look up the technical definition of that, I think you would see something about that it is a science that is building smart machines that can do things that would typically require human intelligence. But I like to flip it around instead of saying what is AI to look at it from the perspective of why do we need AI and why is it important. So if we talk about you have a certain amount of data and we need to look at this data and pick out some trends or information from it. If you have a small amount of data, that is not difficult. So, for example, if you had information that said here are the temperatures in Chicago for 365 days of the year and here our sales of ice cream on those same 365 days. Anybody could easily look at that and say oh, okay, so it looks like when the temperature got warm, we sold a lot of ice cream. When the temperature got cold, we didn’t.
Ed: That is pretty simple to do. Plug the information into Excel, create a graph, and you can immediately see that. But what happens though when you have lots of data hundreds of variables and millions of data points that you are looking at, you quickly get to the point where you have this big data problem where you have so much information that it is impossible for a human to look at that and be able to pick out trends or information from that data. This is where you need a computer to basically take that data and where we can tell it here is what we are looking for. Here is what we want you to find. And let the data or let the computer go in and look at all that data and see what it can pick out from it. The more exciting part of that is machine learning part where if that computer comes back to us with some results that are not what we are looking for that we can actually tell it okay, this is not what we were looking for and here is why. And then the machine gets smarter and then in the future it doesn’t make those mistakes and can basically come up with better information for you.
Ed: That I think is necessary right now in Pharma because in Pharma, we do have a data problem. The number of trials is growing. The number of data points we are collecting is growing. We are getting data now from sources that we never have in the past. Patients wearing a smartwatch and that is supplying data. Patients writing an electronic diary and submitting that. All these wearable devices and things that we are using in trials now that we didn’t in the past, all that is creating massive amounts of data that have to be looked at and obviously analyzed. I think I read something recently that said the amount of data being collected in clinical trials is growing by 40% per year. That is a lot of data, and that is something that is absolutely going to require some type of AI solution.
Jeffrey: Yeah, that is a fantastic explanation and really allows us to visualize, if you will, exactly what the technology is. But I guess now we are at the question is the “so what”, right? So it can look at the data. What does it do with it then, right? So let’s take the practical examples. How can you say it can transform the form of industry by looking at the data? What does that mean? How does that help form over biotech companies?
Ed: Predictive capabilities. I had one person described AI to me as a telescope that allows you to look into the future. And if we can use AI to be able to look ahead and see things that we otherwise would not be able to see, that is going to be basically a huge win for the industry. There are three major problems that clinical trials face. Number one is that they are too expensive. They cost way too much money. Second, they take way too long. But of course the longer a trial takes, the more it is going to end up costing. And then finally it is that we have too many failures. And of course when you start doing a trial you get so far into it, if it fails and doesn’t work out then again that is a lot of time wasted and a lot of expense.
Ed: I have seen predictions that will tell us the cost of a clinical trial averages I think about 2.6 billion dollars. So patients complain about how costly new medicines are. If you perform all these trials and have one drug that makes it to market, you not only have to pay for the research that went into that one. You also have to pay for all those failures and all those drugs that we took to trials that didn’t work out. Therefore, instead of starting that trial, having trouble with the recruitment, canceling it, having it be a failure, we know ahead of time exactly how to conduct this trial before we even get into Phase 1. We know exactly how to run this trial that will give us the highest probability of success, which will lessen the timeline and reduce the cost simply by using AI to look into the future and project these things that we need to know.
Jeffrey: Now that it gave us a better understanding of AI and machine learning, I wanted to speak with someone that is at the forefront of developing AI models for various aspects of drug development. So I spoke with Dr. Olga Kubassova. Now, Dr. Kubassova is a mathematician by training, and then she received her PhD in the area of MRI Algorithm Development. She is now the founding scientist and CEO of IAG. Dr. Kubassova, thank you for joining us today. Can you tell us a little bit about IAG and how your dynamic platform is levering AI and machine learning in drug development.
Dr. Kubassova: IAG is an image analysis group. We came in this industry over 13 years ago, so quite some time ago, with an idea that we are going to bring quantitative image assessment into drug development. What does it mean? It means that we will replace or enhance the way medical images are analyzed by human eye using computer science or mathematics. So we created a platform called dynamical, which incorporates a library of state-of-the-art and AI-driven methodologies.
Jeffrey: To break it down a little bit, can you give some real world example of how this is being used?
Dr. Kubassova: So if you look at the assessment of quite sophisticated MRIs stance, let’s say in solid tumors or neuro-oncology indications such as bloodless tumor. Well, a patient responding to a treatment, it is not a very simple way to respond to treatment. So with advanced therapies, which are currently being developed, the tumor might not necessarily shrink. We would expect with traditional chemotherapies for the tumor to just go down in size and that would be indicative of a patient reacting positively to a therapy. However, today when we develop an immuno-oncology therapies, or you are developing something which will activate your immune system, your tumor may get larger in size when we look at it in the image. However, though it does get larger in size visually, that size does not necessarily represent the tumor itself. It may represent inflammation around the tumor. It is literally impossible for a human eye to look at it and start saying, “Oh, yeah, this is inflammation and this is a tumor.” Because it gets quite sophisticated. So where AI plays a critical role here is really appreciating what each of the image is showing us. It is not recognizable by a human eye, but it is recognizable by a machine and interpretable by a machine. AI helps you to choose the combination which most likely to succeed. The purpose of AI is really that fast, speedy choice and more targeted precise choice. We want to analyze the data related to patient response in a way that human eye can’t possibly see.
Jeffrey: To better understand how AI is being used in the drug discovery process, I spoke with Dr. David Horn Solomon, the CEO of Pharnext. Pharnext is an advanced clinical stage biotechnology company that leverages its artificial intelligence based PLEO therapy platform to develop novel first and class therapies for orphan and common diseases with high unmet needs. Dr. Solomon, thank you for joining us today. Can you tell us about Pharnext PLEO therapy platform and why it is important to your drug discovery program?
Dr. Solomon: Our platform for drug discovery that uses AI really tries to leverage the ideas of Polypharmacology, the idea that multiple medicines often might alter different defective biochemical pathways in diseases and extract a better result for patients and ultimately their families. And so, we are focused on combination medicines that address the complexity of disease. That diseases sometimes caused by single genetic hits but is often caused by defects in multiple biochemical pathways. And so, when you have to sort of these pathways using traditional means, it is often not obvious. And using AI and Big Data, we are able to actually map, if you will, all the genetic pathways, all the biochemical pathways in any given disease given a search of the literature. And then ultimately we are able to use AI to start to map which of these pathways needs tweaking, if you will, and therefore which pathway is targetable and therefore which medicines we can develop against those pathways. And by doing that using AI in a predictive mode, we are really able to start to develop new approaches to medicines that are nonobvious and also innovative.
Jeffrey: Products appears to be reaping the benefits of the platform. The late-stage therapy for Charcot-Marie-Tooth won in the second phase retrial, and you have another candidate in clinical trials for two other neurological indications. So now I think you touched upon it. But how would you say the use of AI has benefited Pharnext? Could a biotech such as yours with limited resources attempt to develop products in three indications like this without the use of AI?
Dr. Solomon: It is much harder because you have to have much larger staff or much more robust high throughput screening to test a lot of the hypotheses that come from understanding the myriad pathways that are defective and how tweaking the right ones can get a good result and therefore cheaper. It is really the speed to get to the clinic that matters because you can spin your wheels in early discovery, not get to the clinic well or rapidly, and that is to the detriment of your shareholder value and ultimately of patients benefit. And so, we think AI here really helps us do this better, faster, cheaper as they might say. And we have done that now not only in Charcot-Marie-Tooth, but as you mentioned also now in Lou Gehrig’s disease and ALS and also to some extent in Alzheimer’s disease.
Jeffrey: Now, it seems like you are focusing on neurological. Is that specific for your AI platform, or is that just areas that you are interested in as a company? Does AI tend to help in one area versus other therapeutic areas?
Dr. Solomon: I think the AI approach here is agnostic to disease or indication. But we also have to marry our discovery platform in our development opportunities with commercial opportunities. What is feasible and what will ultimately benefit patients, especially with AI to think about patients? Whereas there are diseases where there is no current therapy. And obviously ALS is a good example of that. Charcot-Marie-Tooth is clearly there. These patients are on an unstoppable downward decline. When you ask patients, whether it is in New York or Paris or even in the Amazon rainforest, they all say the same thing and that is that today will be the best day of my life because after this it is all downhill. And so, we think that our approach can actually be amenable to diseases where there are no current therapies. And so, that is why we picked those diseases. We also think that there is a significant value for stakeholders or shareholders in these approaches versus cardiovascular disease or diabetes, for example, where there are good existing medicines that are cheap and effective.
Jeffrey: What is next for Pharnext?
Dr. Solomon: Well, keeping along the lines of neurologic disease and again, no promises because we are still in early, early discovery mode. But diseases like myasthenia gravis, they don’t change life expectancy but they make life really miserable for patients and their families, is one area we are very interested in among others. And so we look at the diseases. We look at where our platform is amenable to addressing those diseases and how we might come up with combination medicines. Another opportunity by the way is partnering with other pharma and biotech companies that might have a good medicine, but is not first in class or a leader, a best-in-class. And therefore by adding on adjunct medicines in combination with their existing medicine through perhaps a licensing agreement, we might be able to come up with a best-in-class therapeutic combination in a number of diseases. And so, we are pursuing this approach as well.
Jeffrey: Very interesting. So that is next for Pharnext. Where do you see AI or machine learning going in terms of drug discovery and drug development?
Dr. Solomon: Well, tremendous, tremendous opportunities, just like we are seeing the expansion of AI in other areas. Not only can AI really help figure out how molecules connect to their targets, their receptors with incredible fidelity through a lot of mapping and predictive methods. We think the same is true in AI to develop new medicines. And so, we many examples of these in the literature now and we think as opposed to high throughput screening and other ways that are more laborious on one hand and nonobvious on the other that AI might take us to the promised land, if you will, better, faster, cheaper, simply by doing predictive work. It is like in the game Goer and like in chess, AI allows machines to win. Where here to for, they weren’t able to in the human mind was better in terms of building experience and ability to challenge an opponent. So same is true in drug discovery and we think in the next generation. AI will be a key tool. Again, it is a tool. It is only as good as you apply it. And so, if you apply it to the right diseases with the right thinking with the right molecules with the right targets, I think drug discovery can be greatly aided.
Jeffrey: I would like to thank our guest today Ed Miseta of Clinical Leader, Dr. Olga Kubassova of the Image Analysis Group, and Dr. David Horn Solomon of Pharnext. I would like to thank them for giving us a well-rounded perspective of how artificial intelligence and machine learning is revolutionizing the way new drugs and therapies are coming to market.
Jeffrey: RP HealthCast. Science, Innovation, Life. One story at a time.