An AI expert discusses how close pharmaceutical companies are getting to finding a coronavirus vaccine or treatment and why the use of the technology is helping.
TechRepublic’s Karen Roby talked with Dan Drapeau, an artificial intelligence (AI) expert and head of technology and Blue Fountain Media, about AI’s role in the pharmaceutical industry. The following is an edited transcript of their conversation.
SEE: TechRepublic Premium editorial calendar: IT policies, checklists, toolkits, and research for download (TechRepublic Premium)
Karen Roby: We’re hearing about AI making such an impact in so many different industries, but it might surprise some that it’s really entrenched now in pharmaceuticals.
Dan Drapeau: There’s been a growing need for it. Really, over the past few years, I’d say AI became a little bit more prominent in the pharma space, probably around 2015, 2016, when precision medicine became a little bit more popular. That’s more around coming up with individualized treatment for patients rather than typically what the industry was doing, which is producing big blockbuster-type drugs. So it’s increased in popularity, but now there is more of a true need for it.
Right now with clinical trials, and it’s certainly true with COVID-19 as well, probably one in 10 or so actually make it all the way through the clinical trials process. It costs about a minimum probably of $2 billion to go through that process. It’s a very expensive procedure. Drug discovery takes a long time. So companies are turning to alternative ways to go about reducing that time, reducing operational costs. AI happens to be one of the major solutions that’s become increasingly popular over the last few years.
Karen Roby: Talk a little bit about the specific AI applications, and with COVID-19, how it’s, hopefully, going to make an impact.
Dan Drapeau: At least with COVID-19, this has brought together a large group of professionals from digital technologists, through people in the life sciences space, as well as medical experts. Everyone’s aligned at trying to find a vaccine for this.
AI’s definitely played a role. There’s a large number of companies that are getting closer. I think Moderna is probably the one company that’s closest right now, with a drug being in clinical stage three. But I think one thing that’s in common with all these companies is the use of AI. It’s become a necessity right now, because we’re in the middle of a global pandemic, and it’s vital for us to figure out a solution or vaccine therapy as soon as possible.
SEE: Natural language processing: A cheat sheet (TechRepublic)
AI’s been used in a number of different applications. An interesting way is around repurposing existing drugs. These are drugs that have already been FDA approved. The idea there is cutting time to market. Otherwise, if you come up with a new drug, and it has to go through multiple stages of the clinical trial process, it can take a long time. I think even with Moderna right now, being in phase three for one of their drugs, that started in late July, and it’s going all the way through late October. Then if that’s successful, then the process continues. It’s very time-consuming.
In terms of other applications of AI, well, it really is being used across pharma in so many different ways, from patient targeting for clinical trials to improve success, to drug discovery. I think in general, one of the time-consuming pieces is really around analysis of billions of different molecules and how those might be used to do chemical binding to the target protein that we’re looking at.
Humans can’t possibly do that. It could take forever. They don’t necessarily have the capacity or the time to look at all these different patterns. That’s where AI comes in: It helps automate that process. That’s been a major, significant thing here, as we’re trying to increase the capacity of how best we can go about doing that.
I would say also predicting what drug candidates are most likely to succeed is also pretty important. I mentioned earlier that it’s about $2 billion or so to get a drug all the way through the process. Since about 90% of them don’t succeed, that’s a lot of wasted money. If you have AI that can help predict which ones might be the most successful, that’s a major advantage for some of these pharma companies.
I mentioned looking at existing drugs. I think the whole idea of repurposing drugs is pretty interesting. AI has certainly been used in that capacity.
One other interesting one, not so much related to drug discovery, but it’s become increasingly popular for AI to be used for drug adherence, and figuring out what the correct dosages are. I think that’s a major problem. It’s also a large amount of cost that goes into that sort of thing, so using AI in that capacity has been pretty important as well.
Karen Roby: Dan, I know that you’re in contact with a lot of the big pharma companies. Where are they in this process? You mentioned one of them earlier, but how do they feel about it? Are we getting closer with COVID? Obviously, that’s the question on everybody’s mind right now.
Dan Drapeau: I would say we’re definitely getting closer. There have been huge strides made in the past few months. Specifically as it relates to AI, only about 250 companies, surprisingly, are using AI. I think there’s going to be a major shift after the pandemic is done with companies really realizing that they have to use it moving forward to be successful. But from all the companies I tend to talk with, and that’s a lot of Fortune 500 companies, most are using AI, and a lot of them are working on cures for COVID-19. So it’s already ingrained into their process.
Karen Roby: Dan, any final thoughts as we close out here? It’s, again, fascinating the capabilities of AI. We don’t even know still what machine learning, all of this is capable of down the road.
Dan Drapeau: The most exciting thing for me too, is the evolution of AI with pharma. You just mentioned machine learning. I think one thing to realize here in the whole process is that this is really augmenting the capabilities of scientists who are in the field, not replacing them. I think with machine learning, while we’ve got smaller data sets, it still does require a human to be involved with training of the mathematical model, validating it and so on.
Where everything’s been going a little bit here, especially in the pharma space, is with deep learning. That’s another subtype of AI that uses neural networks. It’s been increasingly important. The benefits there are that it looks at mass amounts of unstructured data. That could be cell images. It could be scientific literature. That actually doesn’t necessarily require scientists or other experts to be involved. It allows, basically with enough data, it can actually figure out what the most important features are itself by effectively self-learning. I think that’s going to be increasingly common here, especially when we’re looking at drug discovery in particular.