A granularity-level information fusion strategy on hypergraph transformer for predicting synergistic effects of...

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Combination therapy shows promise over monotherapy for enhanced treatment efficacy.
• Screening synergistic drug combinations is costly and time-intensive due to the vast number of cancer drugs.
• Existing computational methods for predicting synergistic effects lack exploration of multi-granularity data mining.

Purpose of the Study:

• To propose a novel granularity-level information fusion strategy, HypertranSynergy, for predicting synergistic anticancer drug combinations.
• To address the limitations of existing methods by effectively mining drug and cell line data at different granularity levels.
• To enhance the efficiency and accuracy of identifying effective drug combinations for cancer treatment.

Main Methods:

• Developed HypertranSynergy, a hypergraph transformer model integrating synergistic connections between cancer cell lines and drug combinations.
• Implemented a Coarse-grained Information Extraction (CIE) module with a transformer for node embeddings, featuring Contranorm for over-smoothing mitigation and Gaussian noise for local information.
• Incorporated a Fine-grained Information Extraction (FIE) module utilizing similarity-aware matrices from drug/cell line features to assess fine-grained information impact.

Main Results:

• HypertranSynergy achieved an Area Under the Curve (AUC) of 0.93 ± 0.01 and an Area Under the Precision-Recall Curve (AUPR) of 0.69 ± 0.02 in classification tasks.
• The model demonstrated a Root Mean Squared Error (RMSE) of 13.77 ± 0.07 and a Pearson Correlation Coefficient (PCC) of 0.81 ± 0.02 in regression tasks.
• Performance metrics surpassed those of most state-of-the-art models in predicting synergistic anticancer drug effects.

Conclusions:

• HypertranSynergy effectively integrates multi-granularity information for accurate prediction of synergistic anticancer drug combinations.
• The proposed model offers a significant advancement in computational drug discovery, potentially reducing costs and time in identifying effective combination therapies.
• This approach paves the way for more efficient and precise strategies in personalized cancer treatment.