GRaSP: a graph-based residue neighborhood strategy to predict binding sites

Area of Science:

• Computational biology
• Structural bioinformatics
• Drug discovery

Background:

• Experimental detection of protein-ligand-binding sites is costly and time-consuming.
• In silico methods offer a scalable, fast, and cost-effective alternative for predicting these sites.

Purpose of the Study:

• To introduce GRaSP (Graph-based Residue neighborhood Strategy to Predict binding sites), a novel residue-centric computational method.
• To evaluate GRaSP's performance against existing state-of-the-art methods for predicting ligand-binding sites.

Main Methods:

• GRaSP employs a supervised learning strategy.
• It models the residue environment at the atomic level using a graph-based approach.
• The method is residue-centric and designed for scalability.

Main Results:

• GRaSP demonstrated compatible or superior prediction accuracy compared to existing literature methods.
• It outperformed six other residue-centric methods and achieved better results than a CAMEO-assessed method.
• GRaSP ranked second among five pocket-centric methods, despite not being designed for pocket prediction.
• The method is highly scalable, predicting binding sites in seconds compared to hours for other residue-centric methods.

Conclusions:

• GRaSP represents a significant advancement in predicting protein-ligand-binding sites.
• Its accuracy, speed, and scalability make it a valuable tool for biological research and drug development.
• The open availability of GRaSP facilitates its adoption and further development.