?A large part of what pharmaceutical companies do is research, so they can attempt to design better drugs. Of course, this involves lengthy design and testing processes, but now MIT may have a way to automate at least some of it.
Researchers at MIT¡¯s Computer Science and Artificial Intelligence Lab (CSAIL) have developed a neural network they say is capable of designing drug molecules to an incredibly efficient degree. And if the AI system can be tweaked even further, it may drastically reduce the amount of time it takes in the future to get a new drug from concept to market.
Developing the chemical composition of medication can be a very long drawn out process. Chemists need to build molecules and tweak them to get specific treatments for specific diseases. And sometimes, despite all the money and effort poured into research, you end up with an unstable molecule that¡¯s impossible to manufacture or doesn¡¯t work as planned.
¡°The motivation behind this (AI) was to replace the inefficient human modification process of designing molecules with automated iteration and assure the validity of the molecules we generate,¡± PhD student Wengong Jin from MIT's CSAIL said in a statement. The team trained their AI on 250,000 molecular graphs, i.e. images of a molecule's structure. The bot would then generate molecules, find the best base molecules to build on, and design complex additions to give it medical properties.
In their tests, the AI was able to outperform every other attempt at automated chemical design to come before. When tasked with building completely new molecular structures, each of the AI¡¯s designs turned out to be valid combinations. It was also 30 percent more effective at finding the best base molecule and largely succeeded at modifying molecules to give them more improved medicinal properties than the originals.
Basically, the AI is close to performing with the efficiency of a human chemist and, with some work, could soon outstrip us. Combine that with the fact that the machine works much faster, and we might have robot chemists researching for drug companies very soon.