A fast protein binding site comparison algorithm for proteome-wide protein function prediction and drug repurposing.

Area of Science:

• Structural biology
• Bioinformatics
• Computational chemistry

Background:

• Predicting protein function is crucial for structural genomics.
• Global sequence/structure homology fails for uncharacterized proteins.
• Local ligand binding site similarity offers an alternative for functional inference.

Purpose of the Study:

• Introduce PocketShape, a sequence order-independent algorithm for protein function prediction.
• Assess PocketShape's efficacy in identifying ligand binding site similarity.
• Explore PocketShape's utility in proteome-wide analysis and drug repurposing.

Main Methods:

• Developed PocketShape, a novel algorithm considering backbone geometry, sidechain orientation, and residue properties.
• Evaluated PocketShape on a dataset of 1538 binding site pairs.
• Applied PocketShape to classify 83 diverse enzyme structures and compared performance against existing methods.

Main Results:

• PocketShape achieved 99.3% accuracy in retrieving similar ligand binding site pairs and 100% accuracy in rejecting dissimilar pairs.
• The algorithm successfully classified 83 enzyme structures into 12 clusters, aligning with established protein classifications.
• PocketShape demonstrated superior performance compared to other methods in protein profiling.

Conclusions:

• PocketShape is an accurate and efficient tool for inferring protein function from structural data.
• The algorithm shows promise for proteome-wide function prediction and drug repurposing studies.
• PocketShape identified potential applications for SARS-CoV-2 drugs Remdesivir and 11a.