CFM-DTI: Protein-conditioned feature modulation for drug-target interaction prediction

Area of Science:

• Computational chemistry and cheminformatics
• Bioinformatics and computational biology
• Drug discovery and development

Background:

• Drug-target interaction (DTI) prediction is crucial for drug discovery, but current multimodal methods struggle to capture adaptive target protein influence on drug structures.
• Existing approaches often use simple feature concatenation or weak cross-modal interactions, limiting their ability to model complex relationships.

Purpose of the Study:

• To develop an advanced multimodal DTI prediction framework that effectively integrates drug structural information and target protein context.
• To address the limitations of existing methods by introducing a protein-conditioned asymmetric modulation mechanism.

Main Methods:

• Utilized graph neural networks for drug molecule representation as molecular graphs.
• Employed pretrained ESM2 protein representations for encoding protein sequences.
• Implemented a protein-conditioned feature-wise modulation to adaptively recalibrate drug representations, moving beyond simple symmetric fusion.

Main Results:

• The proposed model achieved superior performance with an AUC of 0.8590 and AUPR of 0.8581 on a dataset of 42,142 drug-target pairs, outperforming established methods like DeepDTA, GraphDTA, and MolTrans.
• Ablation studies confirmed the significant contribution of the molecular graph encoder and the complementary value of pretrained protein features.
• Downstream biological analyses indicated biological coherence and structural plausibility for high-confidence predicted targets.

Conclusions:

• Protein-conditioned asymmetric modulation is an effective strategy for enhancing multimodal DTI prediction by better aligning drug structural features with target semantic context.
• The developed framework offers a promising approach for improving the accuracy and interpretability of DTI prediction in drug discovery.
• Computational analyses provide supportive biological context for prioritized drug targets, guiding further experimental validation.