MegaSyn for Generative Molecule Design
Joshua S. Harris, Fabio Urbina, Sean Ekins · 2024
Even though pharmaceutical companies have increasingly invested in novel biologics such as gene therapies, monoclonal antibodies, and vaccines, small molecule drugs still represent important assets, and they need to find more. The ability to quickly develop novel small molecules for validated targets could offer considerable value to these companies. For many targets of interest to pharmaceutical companies, there is considerable public data available which could be leveraged by machine learning models. Generative molecule design may be a useful tool for enabling the generation of novel small molecules for such targets. We now describe our recent work and improvements to MegaSyn, which is used for generative molecule design. We also provide example applications to demonstrate how MegaSyn can be used to generate molecules predicted to have improved biological and physicochemical properties of interest.