GLA-Synergy: An Interpretable Global-Local Adaptive Framework for Drug Synergy Prediction in Cancer Treatment.

Area of Science:

• Computational biology
• Drug discovery
• Artificial intelligence in medicine

Background:

• Predicting anticancer drug synergy is vital but challenging due to complex biological interactions.
• Existing methods struggle to integrate multimodal features and model hierarchical drug-cell interactions.
• Uncovering the mechanisms of drug synergy is limited by current predictive models.

Purpose of the Study:

• To propose a novel deep learning framework, GLA-Synergy, for enhanced accuracy and interpretability in drug synergy prediction.
• To address limitations in integrating multimodal features and modeling hierarchical interactions.
• To provide an efficient and interpretable tool for discovering synergistic anticancer therapies.

Main Methods:

• GLA-Synergy utilizes a multimodule architecture for global feature extraction from drugs and cell lines.
• An improved linear attention mechanism captures local pairwise interactions.
• Graph convolutional networks, adaptive attention, and dual bilinear attention networks are employed for feature representation and interaction learning.

Main Results:

• GLA-Synergy consistently outperforms existing methods on multiple benchmark datasets.
• The model demonstrates improved accuracy in predicting drug synergy.
• The framework provides enhanced interpretability of synergistic interactions.

Conclusions:

• GLA-Synergy offers a powerful and interpretable approach to predicting anticancer drug synergy.
• The progressive global-local fusion framework effectively captures critical drug-cell interactions.
• This model facilitates the discovery of novel synergistic anticancer drug combinations.