Recursion Pharmaceuticals, Inc.

Cyclica Inc.

Stephen Scott Mackinnon

Faud G Gwadry

Daniella Fiora Lato

Andreas Windemuth

Seyed Ali Madani Tonekaboni

Andrew Edward Brereton

Sana Alwash

Joseph Christian Campbell Somody

Fuad G Gwadry

Jason Mitakidis

Leonard David Morayniss

The present disclosure relates to systems, non-transitory computer-readable media, and methods that utilizing compound-protein machine learning representations to generate target results. For example, the disclosed systems can utilize a compound-protein interaction machine learning model to generate a compound-protein machine learning representation for compound protein pairs. The disclosed systems can utilize the compound-protein machine learning representation to train and utilize other target machine learning models in generating predicted bioactivity results. For example, the disclosed systems train a target machine learning model from compound-protein machine learning representations to generate ADMET predictions and/or biological perturbation program predictions. Furthermore, the disclosed systems can utilize one or more explainability models in conjunction with target machine learning models trained based on compound-protein machine learning representations to identify proteins that contribute to predicted bioactivity results.

Utilizing compound-protein machine learning representations to generate bioactivity predictions

A method for predicting a property of a sample molecule involves, for each of a multitude of reference molecules, obtaining a multitude of fingerprints and at least one property, and obtaining the multitude of fingerprints of the sample molecule. The method further involves for each of the multitude of reference molecules, using each of the multitude of fingerprints, calculating distances to the sample molecule, and for each of the multitude of reference molecules, determining a relative predictive dominance, based on the distances to the sample molecule. The method also involves, for each of the multitude of reference molecules, determining a fitness value based on the relative predictive dominance, and predicting the at least one property of the sample molecule based on the at least one property of the multitude of reference molecules and the fitness values obtained for the reference molecules.

Method and system for predicting properties of chemical structures

The invention involves a method for identifying a target protein. The invention involves receiving a request to identify a target protein based on a ligand; identifying, using the ligand, a first protein, where the ligand binds with the first protein to form a ligand-protein complex; generating, a first binding site profile for the first protein, where the first binding site profile describes molecular properties of the first protein; obtaining, from a controlled server, structure data describing molecular properties of surfaces for a multitude of proteins, where the multitude of proteins comprises the target protein; identifying, using the first binding site profile and the structure data, the target protein; and presenting the target protein to a user.

Methods for proteome docking to identify protein-ligand interactions

Growing community of inventors

Burlington, Canada

Stephen Scott Mackinnon