In this interview with James W. Baurley, co-founder of BioRealm, we discuss some of the benefits we can expect from AI in genomic research, AI’s role in solving genomic research problems, why researchers should be considering AI for their projects, and how they should get started.
Estimated Read Time: ~4 minutes
Conor Ryan: We can see the benefits of artificial intelligence (AI) replacing human drivers pretty easily. Would you give us a few examples of the benefits we can expect from AI in genomic research?
James W. Baurley: Sure. Let’s start with human disease and agriculture. AI can help us learn who’s at a higher risk for specific diseases, and which treatments will give them the best results with the fewest side effects. These are very important questions because we can convert those answers into real-world solutions to help people.
An example is the work we’ve been doing to determine who’s more likely to become addicted to smoking tobacco and which treatment(s) will get them to stop smoking as quickly and easily as possible with the fewest side effects. One possible future benefit of that might be a simple testing kit you could use at home that would tell you the best treatment(s) available for you to stop smoking.
CR: And agriculture?
JWB: Earlier this year we constructed computer models for a rice research program managed by the United Nations. Rice has about 45,000 naturally occurring varieties, and over 100,000 cultivated varieties which are the results of traditional plant breeding. Our models predicted the best known varieties of rice to plant in given field conditions. Those types of models could also tell us which varieties we should breed to create a new cultivated variety that has the desired combination of characteristics, like disease and drought resistance, yield, and taste.
CR: In autonomous vehicles, AI replaces the driver, what is AI’s role in solving these problems?
JWB: AI helps us identify the parts of the genome that are relevant to different biological processes and traits, disease prediction, and so on. We can’t build computer models without AI because there’s too much data and it’s too complex. For example, with plants, some data comes from things we can measure, like how fast a plant is growing; some data are genetic, like disease resistance; and some data are non-genetic, like weather. That’s a lot of data, especially the genetic data, and the problem usually gets much bigger when we start working with the human genome. We need to discover and model the relationships within all of that data. AI can make sense of many variables, and even learn new variables, where humans simply can’t.
CR: Why is this possible now?
JWB: We need a lot of data to see all of these relationships. Until recently, we usually didn’t have enough data, and even if we did we didn’t have the computer processing power, storage, and algorithms to take full advantage of it all. We’ve seen huge improvements in computer hardware and software, and individual genomic data is much more readily available—we even have direct-to-consumer genetic testing kits. All of this made it possible to actually use some algorithms that were decades old, like neural networks, and to create new algorithms, like layered-neural networks, also known as deep learning.
CR: Last year Dr. Carolyn Ervin, a co-founder of BioRealm, shared some anti-patterns when it comes to how people use software in biostatistics. Do you have similar concerns when it comes to genomic research?
JWB: The biggest challenge facing our clients today is the same as it was more than ten years ago when we started BioRealm—all the things they don’t know they need to know. Today’s projects need expertise in many fields to succeed, and increasingly that includes expertise in AI. For example, if you want to work on smoking cessation, you can’t just fill your team with experts on smoking cessation. To do serious work you’ll need experts in computer hardware and software, including AI, as well as experts in epidemiology, regulations, statistics, study design, and so on.
CR: Why is that so important?
JWB: Because in every field you need to make sure you’re taking full advantage of the latest advances, and, more importantly, that you’ve got all the basics covered. Projects need to start with a solid foundation. The only way to do that is to have a diverse team of experts. When people ignore the expertise required in other fields, they put their projects in danger. We’re called in to rescue projects all the time because of this.
CR: How can readers make sure they’ve got all the resources they need?
JWB: Reach out and get help! Either hire us to do a preliminary evaluation, or assemble a team of experts yourself. Whatever you do, get help as early as possible. Catching problems early makes your life so much easier, and saves a lot of time, money, and stress.
CR: What are some other challenges you’re seeing when it comes to clients attempting to use AI in genomic research?
JWB: AI may be the special sauce needed to kick-start a project, and it’s great when we get to help clients get those benefits. They need to understand that there are limitations to AI, that sometimes it isn’t beneficial, and that they need to address the fundamentals first, then determine if AI might help.