Reproduce collective knowledge to aid drug discovery. Artificial intelligence can use human "chemical intuition"

A paper published in Nature Communications on the 31st reported a machine learning model that can partially reproduce the collective knowledge accumulated by professional chemists in their work. This type of knowledge is often called "chemical intuition." The research team believes that this research may make future drug development more efficient.

Overall research framework image source: "Nature Communications" online version

Traditionally, drug and chemical discovery has relied on trial-and-error experiments and knowledge accumulated by researchers on the job. Using simulation tools, especially machine learning, allows researchers to discover candidate molecules faster, greatly reducing the cost of discovering new pharmaceutical compounds. If machine learning is to be used to predict molecular properties, the molecule must be reduced to a mathematical representation, which typically consists of a set of properties or "features." Identifying the right features is key to the success of these data-driven performance prediction models.

A joint team from the Microsoft Research Center for Science Intelligence (AI4Science) in the UK and the Novartis Institutes for Biomedical Research in Switzerland asked 35 medical chemists to each choose their preferred molecule from 5,000 pairs of molecules, and then used their answers to Play a ranking game to train a machine learning model and then have the model score molecules. This score comes from years of accumulated knowledge in the industry and is basically unaffected by other factors.

The model proposed by the team can also be used to alter the recommendations of mathematical models to better match the collective expertise of chemists, potentially shortening iteration times in future early-stage drug development.

The researchers believe this method may be useful as a complement to molecular modeling in drug development.

In the settings of some science fiction novels, alien creatures are born with all the accumulated knowledge of the current era, thereby achieving rapid development of civilization. Human beings can only master knowledge and accumulate experience by studying for a long time. This time, the machine learned intuition, which is equivalent to transplanting valuable human experience in an instant, just like Alpha Dog learned quickly and instantly defeated the human master in the field of Go. AI is revolutionizing chemistry. It promotes rapid progress in precision chemistry in terms of new material discovery, accurate simulation, and optimization of synthetic pathways, and can also significantly shorten the trial-and-error process of drug development.