SAFER: sub-hypergraph attention-based neural network for predicting effective responses to dose combinations

Area of Science:

• Computational biology
• Pharmacology
• Artificial intelligence

Background:

• Drug combination synergy offers significant benefits in cancer treatment but poses risks of increased toxicity.
• Current artificial intelligence (AI) models for predicting drug synergy often overlook crucial dosage information and dynamic biological networks, limiting their applicability.
• Existing graph-based models typically use static protein-protein interactions, failing to capture context-dependent biological network dynamics.

Approach:

• Introduced SAFER (Sub-hypergraph Attention-based graph model), an AI framework designed to predict drug combination synergy.
• SAFER incorporates complex biological knowledge networks and considers the impact of dosing on subject-specific networks.
• The model utilizes a sub-hypergraph attention mechanism to analyze relationships within biological networks.

Key Points:

• SAFER accurately predicts drug combination synergy, outperforming previous models on benchmark and independent test datasets.
• The model's analysis identified the JAK-STAT signaling pathway, PRDM12, ZNF781, and CDC5L as relevant to lung cancer, with implications for lung fibrosis.
• SAFER provides an interpretable framework for identifying drug-responsive signals and understanding dose-level responses.

Conclusions:

• SAFER offers a novel, interpretable framework for predicting dose-dependent drug combination synergy in cancer therapy.
• The model's ability to capture subject-specific molecular contexts and dose effects opens new research avenues.
• SAFER can be utilized in future studies to investigate patient-specific molecular networks for personalized medicine.