Disentangling Causal Substructures for Interpretable and Generalizable Drug Synergy Prediction

2025-11-04

AI Summary

Existing drug synergy prediction models are predominantly black-box, relying on statistical correlations without elucidating underlying molecular mechanisms; they also suffer from poor generalizability in cold-start and out-of-distribution settings. This paper proposes CausalDDS, the first framework to integrate causal representation learning into drug synergy prediction. It employs conditional intervention modeling to disentangle causal substructures—those genuinely driving synergy—from spurious or redundant ones within drug molecules, jointly optimizing representation separation via sufficiency and independence constraints. CausalDDS significantly improves both predictive accuracy and interpretability, outperforming state-of-the-art methods across multiple benchmarks. Notably, it demonstrates robust generalization under cold-start and distributional shift conditions. Moreover, it enables precise identification of pharmacophore substructures responsible for synergy, establishing a new paradigm for trustworthy, mechanism-aware drug discovery.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Drug synergy prediction is a critical task in the development of effective combination therapies for complex diseases, including cancer. Although existing methods have shown promising results, they often operate as black-box predictors that rely predominantly on statistical correlations between drug characteristics and results. To address this limitation, we propose CausalDDS, a novel framework that disentangles drug molecules into causal and spurious substructures, utilizing the causal substructure representations for predicting drug synergy. By focusing on causal sub-structures, CausalDDS effectively mitigates the impact of redundant features introduced by spurious substructures, enhancing the accuracy and interpretability of the model. In addition, CausalDDS employs a conditional intervention mechanism, where interventions are conditioned on paired molecular structures, and introduces a novel optimization objective guided by the principles of sufficiency and independence. Extensive experiments demonstrate that our method outperforms baseline models, particularly in cold start and out-of-distribution settings. Besides, CausalDDS effectively identifies key substructures underlying drug synergy, providing clear insights into how drug combinations work at the molecular level. These results underscore the potential of CausalDDS as a practical tool for predicting drug synergy and facilitating drug discovery.

Problem

• Predicting drug synergy by identifying causal molecular substructures
• Enhancing interpretability and generalizability in drug combination prediction
• Mitigating spurious correlations from redundant molecular features

Innovation

• Disentangles drug molecules into causal and spurious substructures
• Employs conditional intervention mechanism on paired structures
• Introduces sufficiency and independence guided optimization objective