Multi-parameter optimization (MPO) is a major challenge confronting the drug discovery and development process, making it difficult to identify promising molecules meeting the Target Product Profile (TPP), thus making it an expensive, complex, and time-consuming task. Recently, the development of artificial intelligence (AI) approaches to de novo drug design through the use of deep generative models has triggered significant interest in the drug hunter community as an important tool to accelerate the process. To that end, we developed Makya (makya.ai), a chemist-friendly platform for generative AI-driven de novo drug design to (i) enhance the explored chemical space while tackling the MPO challenge, and (ii) better address the chemical feasibility of generated compounds. Makya uses deep-learning based de novo design algorithms to train QSAR models. The molecular generation process is optimized by Reinforcement Learning based on the diversity, novelty, quality, and synthetic accessibility of the generated compounds. Here we present the results of Makya implementation to drug discovery projects (from hit discovery to lead optimization) both in industry and academia. Use of Makya in these projects has led to identification of new chemotypes, filing of patent applications, and resolution of project specific issues (selectivity, activity, ADME, etc). This has also resulted in the first reported use of AI-based generative modeling to solve the MPO problem by identifying a molecule that met all objectives of the TPP. We demonstrate the benefit of applying AI technology to accelerate real drug discovery projects.

Makya: AI-driven de novo drug design platform focused on multi-parametric optimization

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.