When Deep Learning and Drug Discovery Intersect
Insilico Medicine has announced the publication of a special issue of Molecular Pharmaceutics, run by the American Chemical Society, entitled: “Deep learning for drug discovery and biomarker development”. The magazine, which includes an article written by Insilico’s CEO, will be dedicated to recent applications of deep learning in drug discovery and biomarker development, and will bring together contributions from top academics and experts in the field, particularly regarding generative chemistry and the related generative adversarial networks (GANs).
Deep learning as a concept began gaining ground in 2014 as artificial intelligence began to outperform humans in areas ranging from playing video games to autonomous driving. Images generated by GANs from natural language began to appear at around the same time.
Despite this, it is only recently that deep learning has become prevalent in the fields of generative chemistry and GANs, largely due to lengthy validation cycles and the working gap between chemists, biologists and AI scientists.
The special issue of Molecular Pharmaceutics will largely focus on generative chemistry using GANs and reinforcement learning for de novo molecular design. Among other things the articles demonstrate for the first time the experimental validation of the molecules generated using these architectures.
The issue announces among other things that entangled conditional adversarial autoencoder was used to generate a novel inhibitor of Janus Kinase 3, which is thought to play a role in rheumatoid arthritis and psoriasis, among other diseases. When the molecule was tested in vitro, it showed high activity and selectivity.