Discovering Novel Pharmaceutical Molecules with Generative Adversarial Networks
Leena Chandrashekar, K. Namratha, P.M. Mohan, Raxna Bojamma P J, Aditi Kulkarni · 2023
The purpose of drug design is to construct a chemical component that can fit into a specific cavity on a protein target both physically and chemically. Traditional drug design procedures are both costly and time demanding. The era of computer-aided drug design (CADD), provided a quantitative relationship between structure and biological activity. However, due to time constraints and a shortage of trained and knowledgeable pharmaceutical personnel, it fell short of precision. With the increasing popularity of AI-based methods, the pharmaceutical industry also gained acceleration and streamlined the drug development pipeline. Generative Adversarial Networks(GAN) offered very good results in the creation of new drugs. This paper discusses the design and implementation of GAN with experiments on the QM9 dataset. The molecules produced are evaluated with various performance metrics and also compared with benchmark models such as MolGAN and ORGAN.