A graph attention-based deep learning network for predicting biotech-small-molecule drug interactions.

Abstract

Motivation: The increasing demand for effective drug combinations has made drug-drug interaction prediction a critical task in modern pharmacology. While most existing research focuses on small-molecule drugs, the role of biotech drugs in complex disease treatments remains relatively unexplored. Biotech drugs, derived from biological sources, have unique molecular structures that differ significantly from those of small molecules, making their interactions more challenging to predict.

Results: This study introduces a novel graph attention network-based deep learning framework that improves interaction prediction between biotech and small-molecule drugs. Experimental results demonstrate that the proposed method outperforms existing methods in multiclass drug-drug interaction prediction, achieving superior performance across various evaluation types, including micro, macro, and weighted assessments. These findings highlight the potential of deep learning and graph-based models in uncovering novel interactions between biotech and small-molecule drugs, paving the way for more effective combination therapies in drug discovery.

Availability and implementation: The datasets and source code of this study are available in the GitHub repository: https://github.com/BioinformaticsIASBS/BSI-Net.