RECOVER identifies synergistic drug combinations in vitro through sequential model optimization

  • Cell Rep Methods. 2023 Oct 23;3(10):100599. doi: 10.1016/j.crmeth.2023.100599.
Paul Bertin  1 ,  Jarrid Rector-Brooks  1 ,  Deepak Sharma  1 ,  Thomas Gaudelet  2 ,  Andrew Anighoro  2 ,  Torsten Gross  2 ,  Francisco Martínez-Peña  3 ,  Eileen L Tang  3 ,  M S Suraj  2 ,  Cristian Regep  2 ,  Jeremy B R Hayter  2 ,  Maksym Korablyov  1 ,  Nicholas Valiante  4 ,  Almer van der Sloot  5 ,  Mike Tyers  6 ,  Charles E S Roberts  2 ,  Michael M Bronstein  7 ,  Luke L Lairson  3 ,  Jake P Taylor-King  8 ,  Yoshua Bengio  1
Affiliations
  • 1. Mila, the Quebec AI Institute, Montreal, QC, Canada.
  • 2. Relation Therapeutics, London, UK.
  • 3. Department of Chemistry, The Scripps Research Institute, La Jolla, CA, USA.
  • 4. Innovac Therapeutics, Inc., Cambridge, MA, USA.
  • 5. IRIC, Institute for Research in Immunology and Cancer, Université de Montréal, Montreal, QC, Canada.
  • 6. Program in Molecular Medicine, Peter Gilgan Centre for Research and Learning, The Hospital for Sick Children, 686 Bay Street, Toronto, ON M5G 0A4, Canada.
  • 7. Relation Therapeutics, London, UK; Department of Computer Science, University of Oxford, Oxford, UK.
  • 8. Relation Therapeutics, London, UK. Electronic address: [email protected].
PMID: 37797618 DOI: 10.1016/j.crmeth.2023.100599
Abstract

For large libraries of small molecules, exhaustive combinatorial chemical screens become infeasible to perform when considering a range of disease models, assay conditions, and dose ranges. Deep learning models have achieved state-of-the-art results in silico for the prediction of synergy scores. However, databases of drug combinations are biased toward synergistic agents and results do not generalize out of distribution. During 5 rounds of experimentation, we employ sequential model optimization with a deep learning model to select drug combinations increasingly enriched for synergism and active against a Cancer cell line-evaluating only ∼5% of the total search space. Moreover, we find that learned drug embeddings (using structural information) begin to reflect biological mechanisms. In silico benchmarking suggests search queries are ∼5-10× enriched for highly synergistic drug combinations by using sequential rounds of evaluation when compared with random selection or ∼3× when using a pretrained model.

Keywords
CP: Systems biology; active learning; deep learning; drug combination; drug synergy; in vitro screening; machine learning; oncology; sequential model optimization.
Products