Knowledge-Aware Synergistic Discovery of Drug Combinations: A Large Language Model Perspective.

Area of Science:

• Oncology
• Computational Biology
• Pharmacology

Background:

• Drug combination therapy is a cornerstone of cancer treatment, offering synergistic advantages.
• Current deep learning methods for drug synergy prediction often overlook domain-specific data characteristics and knowledge integration.
• Effectively incorporating dispersed professional knowledge into data mining remains a significant challenge.

Purpose of the Study:

• To propose KSDDC, a novel knowledge-aware model for synergistic drug combination discovery.
• To leverage large language models (LLMs) for enhanced drug synergy prediction by integrating systematic knowledge.
• To improve the accuracy and reliability of predicting effective drug combinations in cancer therapy.

Main Methods:

• Developed KSDDC, a model utilizing a large language model (LLM) perspective for knowledge integration.
• Implemented three core modules: knowledge-aware drug feature auto-encoding, knowledge-aware cell line feature encoding, and drug-drug synergy prediction.
• Generated informative embeddings by combining sample features for accurate synergy prediction.

Main Results:

• KSDDC demonstrated superior performance compared to shallow and deep machine learning methods on synergy prediction benchmarks.
• Achieved approximately 19% F1-score improvement over the second-best method on the DrugComb_1 dataset.
• Validated the effectiveness of knowledge-enabled data mining in drug synergy discovery.

Conclusions:

• KSDDC offers a novel and effective approach for knowledge-aware synergistic drug discovery.
• The model's ability to integrate professional knowledge enhances drug synergy prediction accuracy.
• This study provides valuable insights and a reference method for future research in cancer drug combination therapies.