Accurate Identification of Protein Binding Sites for All Drug Modalities Using ALLSites.

Area of Science:

• Computational biology
• Drug discovery
• Bioinformatics

Background:

• Identifying protein binding sites is crucial for drug discovery but current methods are limited.
• Existing computational tools often lack specificity, require structural data, or have insufficient accuracy.

Purpose of the Study:

• To develop a unified, sequence-based framework for predicting proteome-wide binding sites across all drug modalities.
• To overcome the limitations of existing binding site prediction methods.

Main Methods:

• Developed ALLSites, a sequence-based framework utilizing ESM-2 embeddings.
• Integrated a gated convolutional network with a transformer architecture to capture global and local sequence features.
• Modeled residue interactions directly from protein sequence data.

Main Results:

• ALLSites achieves superior predictive performance across diverse drug modalities (proteins, peptides, small molecules, carbohydrates, DNA, RNA).
• Demonstrates state-of-the-art performance among sequence-based methods.
• Matches the accuracy of the best structure-based prediction tools.

Conclusions:

• ALLSites enables accurate, structure-free binding site prediction for all drug modalities.
• Expected to significantly expand the druggable proteome.
• Provides a powerful new resource for accelerating drug discovery efforts.