Where is AI-Driven Bioengineering Headed Next?

Everybody knows that AI is good at detecting patterns. This can be very useful when looking at electronic medical records for finding a bunch of patients whose disease looks similar. A lot of diseases are really waste bins of all different kinds of people with slightly different diseases. When you do a clinical trial, of course your drug is only going to work on 20% of the people because 20% of them actually have a version of the disease where that drug is relevant. Percent might not even have the disease, or they just have a totally different form. The AI systems are very good at finding patients who are kind of looking very similar along all available dimensions, and that's also very valuable. Now jumping very molecular We've all heard about AlphaFold2. AlphaFold2 is the program out of DeepMind that was able to predict the three dimensional structure of proteins. I'll just remind people who haven't taken biology since high school that those three dimensional proteins are typically what we call the targets of a drug. A drug is often a small molecule that binds one of these proteins and modulates its function to help the patient have their disease go away or get better. Going from just some protein 3 dimensional structures to all of the structures, at least almost all at a reasonable level of accuracy, opens up our ability to think about in a very rational way, new drugs to interact with proteins that we weren't able to think about previously because we didn't have the structure. And so that whole three dimensional structure and molecular understanding of drug action is about to be revolutionized. I mean, it's happening right now. I can't tell you how quickly engineering students read those papers, walked into my office and said, I want to work on the spin out effects of that discovery.