In this inaugural DREAM Target 2035 Challenge, we are inviting the community of data scientists to build Machine learning models based on molecular data from DNA encoded libraries (or DELs) to efficiently discover hits (bioactive drug-like molecules) This DEL-ML approach was introduced in 2020 by a team at Google and X-Chem.
Goal:
- Build and train Machine Learning models on data from a DNA-encoded library (DEL) screen focused on WDR91, a target protein, and retrospectively retrieve 145 confirmed hits hidden in an unrelated chemical library of ~370,000 diverse molecules.
- if you pass step #1, use your validated model to prospectively predict novel hits from a commercial library of 4.4 million compounds. Your predicted molecules will be tested experimentally at the Structural Genomics Consortium (SGC) , University of Toronto, and you will be able to publish your discovery. The winner will receive a $5,000 cash prize.
Motivation:
Target 2035 is an open science global movement consisting of international scientists and researchers, focusing on the creation of chemical and biological tools to study human proteins and inform drug discovery.
The success of Target 2035 relies on future breakthroughs in machine learning (ML) for drug discovery. To that end, the SGC and its industry partners are populating the AIRCHECK platform with training data and (soon) best performing ML models.
