The Upsides of Machine Learning in Drug Discovery

A colleague of mine has been looking at the automated discovery of what are called small molecules in chemical engineering and chemistry. The problem with it was finding molecules that had the desired properties but were more complicated to actually synthesize than a human chemist or engineer would have come up with. He used basically a chemical catalog of simpler compounds that you use in the lab to do synthesis. And by adding that to the training process, he enabled the AI to not only find these things that have properties where people haven't been able to find reasonable compounds yet, but ones that were vastly simpler to synthesize. If this kind of work, I think where machine learning can really drive things.