HLA-EpiCheck: novel approach for HLA B-cell epitope prediction using 3D-surface patch descriptors derived from...

Area of Science:

• Immunology and Transplantation Science
• Computational Biology and Bioinformatics
• Structural Biology

Background:

• The human leukocyte antigen (HLA) system is a primary driver of organ transplant rejection due to recipient-generated donor-specific antibodies recognizing graft HLAs.
• Identifying immunogenic B-cell epitopes on HLAs is essential for improving organ allocation and reducing transplant loss.
• Current methods for characterizing HLA epitopes, based on polymorphic residues called 'eplets', have limitations, with many polymorphic positions unconfirmed and structural studies often neglecting dynamic aspects.

Purpose of the Study:

• To develop a novel machine learning approach for predicting immunogenic B-cell epitopes on human leukocyte antigens (HLAs).
• To incorporate dynamic structural information from molecular dynamics simulations into epitope prediction.
• To improve the accuracy and scope of HLA epitope identification beyond current experimental and static structural limitations.

Main Methods:

• A machine learning framework, HLA-EpiCheck, was developed using 3D-surface patches derived from 207 HLA structures (solved and predicted).
• Epitope and non-epitope patches were labeled using data from the Human Leukocyte Antigen Eplet Registry.
• The system utilizes descriptors from both static and dynamic properties of 3D-surface patches, with tree-based models trained on a non-redundant dataset.

Main Results:

• The HLA-EpiCheck system demonstrated strong performance, leveraging dynamic descriptors for over half of its predictive power.
• Predictions were made for unconfirmed eplets not initially included in the training dataset.
• A notable consistency was observed when comparing HLA-EpiCheck's predictions with experimental results for these novel eplets.

Conclusions:

• The developed machine learning approach effectively predicts HLA epitopes by integrating dynamic structural information.
• HLA-EpiCheck offers a promising tool for refining HLA epitope identification, potentially improving organ allocation strategies.
• The study highlights the importance of molecular dynamics simulations in understanding and predicting HLA immunogenicity.