Drug Discovery

Drug discovery is an important area with big business and social impact. From designing a new drug until it is finally approved, it takes tens of years and billions of dollars. We would like to apply machine learning techniques to this area and speed up the process.

Before conducting wet experiments and clinical trials, there are several steps that machine learning can help:

1. Target validation, which is to mine possible targets from litearture, proteomics, etc;
2. Screening, which is about to search libraries to find possible molecules that are most like to bind with drug targers;
3. Lead generation/optimization, which is to improve the potency of the selected molecules.

Our research directions cover the above areas, including literature mining, molecule pre-training, drug property prediction, drug-target interaction prediction, new generative models and retrosynthesis.