It has been recently estimated that, on average, the process of bringing a single new drug to market takes 8.3 years and costs $1,336 million. Yet, only about 2.4% of the drugs that enter clinical trials reach FDA approval. Structure-based drug design can make an impact on this approval rate, by helping develop more potent and selective drugs. Here we present Elion, a complete AI-based drug design platform, which forms an end-to-end iterative cycle including a reinforcement-based molecular library generator, ADMET predictors, docking, and a deep learning method for precise estimation of binding energies. As an example, we apply this workflow to generate new protein-protein interaction inhibitors for the YAP/TEAD transcriptional system. TEAD proteins are the final effectors of the Hippo pathway, an important regulator of organ size control, cell growth, and proliferation, and require binding to the transcriptional co-activator YAP or TAZ. TEAD expression is upregulated in many cancer types including gastric, colorectal, pancreatic, liver, breast, and prostate cancer, and is associated with drug resistance and low patient survival. Disrupting the YAP-TEAD interaction is a promising therapeutic strategy against cancers. The drug design process starts with a virtual high-throughput screening of a database with 100,000 diverse molecules, which provides the initial hit and activity model. This model is used, together with other properties such as drug-likeness and synthetic accessibility, to train a reinforcement-learning-based molecular generator in creating new molecules with a similar scaffold to the hit molecule while optimizing the desired properties. New molecules are then docked into the target, and the poses are rescored using DeepAtom, a method for calculation of binding energies which uses a 3D-convolutional neural network. As a result, we suggest a set of molecules that are potentially more active and less toxic than the original hit, which will be synthesized and tested in our laboratory.

Elion: An artificial intelligence-based computational platform for drug discovery and optimization

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

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