HLA-Clus: HLA class I clustering based on 3D structure.

Area of Science:

• Immunogenetics
• Structural Biology
• Bioinformatics

Background:

• Previous work classified HLA Class I alleles into supertypes/subtypes using 3D peptide binding groove similarities.
• The established method demonstrated high correlation with peptide binding specificity, cohesion, and robustness.

Purpose of the Study:

• Introduce HLA-Clus, a Python package for clustering HLA Class I alleles.
• Incorporate a novel nearest neighbor clustering method for user-defined criteria.

Main Methods:

• Coarse-graining HLA Class I structural models into labeled point clouds.
• Calculating allele similarities using a structure distance metric capturing spatial and physicochemical properties.
• Clustering alleles via hierarchical or nearest-neighbor approaches.

Main Results:

• HLA-Clus pipeline successfully clusters HLA Class I alleles.
• Integration with MHCnuggets improved peptide binding prediction accuracy for rare alleles via nearest neighbor selection.
• The package facilitates characterization of peptide binding specificities for numerous HLA alleles.

Conclusions:

• HLA-Clus provides a robust solution for HLA allele characterization.
• Applications include peptide affinity prediction, disease association studies, and HLA matching for transplantation.
• The HLA-Clus package is publicly available on GitHub.