Drug Target Prediction from Perturbation Transcriptomics via a Biological Function-Guided Hypergraph Siamese Network

Area of Science:

• Computational biology
• Genomics
• Drug discovery

Background:

• Elucidating small molecule effects on cellular states is crucial for drug discovery.
• Perturbation transcriptomics provides insights into drug-target interactions by analyzing gene expression changes.

Purpose of the Study:

• To develop BioHSNet, a biological function-guided hypergraph siamese network for inferring drug-target interactions.
• To leverage gene expression data and compound structures for improved interaction prediction.

Main Methods:

• BioHSNet employs hyperedge representations of functionally grouped gene expression.
• It captures higher-order functional relationships and integrates compound structural information.
• The model was evaluated on the Broad Institute's L1000 datasets.

Main Results:

• BioHSNet outperforms existing transcriptome-based methods in inferring drug-target interactions.
• The model demonstrates superior performance in cold start scenarios.
• A case study highlighted its utility in target prediction and drug screening.

Conclusions:

• BioHSNet offers a powerful approach for drug-target interaction inference.
• Its ability to integrate diverse data types enhances prediction accuracy.
• The method shows promise for accelerating drug discovery and development.