Cell Reports: Methods (Oct 2023)
RECOVER identifies synergistic drug combinations in vitro through sequential model optimization
Abstract
Summary: 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. Motivation: Galvanized by the COVID-19 pandemic, we wanted to systematically identify efficacious drug combinations from the plethora of safe drugs that could hypothetically exhibit antiviral activity. The infeasibility of extensive combinatorial screens triggered the need for new methods that would require substantially less screening than an exhaustive evaluation. Outside of biology, there has been much interest in how areas of machine learning, including active learning and sequential model optimization, can be utilized to efficiently explore large spaces of possibilities through the intelligent acquisition and interpretation of data. Sequential model optimization has received much interest within biomedicine, with a focus on systems with well-described individual components, e.g., biomolecular design, chemical assays, etc. We wanted to apply a similar philosophy to quickly identify synergistic drug combinations to alter the phenotype of a cellular model system (cell viability as proof of concept), where the relationship between the chemical inputs and resulting phenotypic output is not well understood and is subject to experimental biases.
