Ádám Arany

Research

Data integration and drug-target interaction prediction for the pharmaceutical industry

The research goal is to generate testable hypotheses based on an industrial scale dataset with millions of compounds and thousands of protein targets with 100s of millions of measured activity values. We should be able to derive reliable and well calibrated credibility intervals for our hypotheses to make the following decision making process possible. We develop and apply a Bayesian matrix factorization method Macau to make predictions in the incompletely filled drug-protein matrix, using different type of side-informations, e.g.: chemical structure fingerprints, protein target similarities, imaging-based High-Content Screening (HCS) results.

Active projects

• AIDD - Advanced machine learning for Innovative Drug Discovery

Finished projects

• CELSA - Active Learning - Active learning and interpretation of computational model on tissue specific permeability for stock and virtual drug discovery library
• ChemBioBridge - Bridging chemical and biological experimental design for pharma industry through data fusion.
• Melloddy - MachinE Learning Ledger Orchestration for Drug DiscoverY

Publications