HLA amino acid Mismatch-Based risk stratification of kidney allograft failure using a novel Machine learning...

Area of Science:

• Immunogenetics
• Transplantation immunology
• Computational biology

Background:

• Human Leukocyte Antigen (HLA) antigen-level mismatches (Ag-MM) are linked to kidney allograft failure.
• Amino acid-level mismatches (AA-MM) offer a more granular view but are less explored.
• Ag-MM may obscure the variable impact of mismatches at polymorphic sites.

Purpose of the Study:

• Develop a novel algorithm, Feature Inclusion Bin Evolver for Risk Stratification (FIBERS).
• Apply FIBERS to discover HLA AA-MM bins for stratifying kidney transplant donor-recipient pairs into low vs. high graft survival risk groups.
• Compare FIBERS's performance against traditional Ag-MM risk stratification.

Main Methods:

• Applied FIBERS to 166,574 kidney transplants from 2000-2017 (Scientific Registry of Transplant Recipients).
• Evaluated AA-MMs across all HLA loci (A, B, C, DRB1, DQB1) and individually.
• Used cross-validation and adjusted Cox models for predictive power assessment, controlling for covariates and Ag-MMs.

Main Results:

• FIBERS's best bin significantly improved graft failure risk stratification (HR=1.10, p<0.001), identifying more low-risk patients than 0-ABDR Ag-MM.
• HLA-DRB1 AA-MMs, particularly at peptide contact sites, showed the strongest individual locus risk stratification (HR=1.11, p<0.005).
• Potential risks were also identified for HLA-DQB1 AA-MMs impacting peptide binding and heterodimer stability.

Conclusions:

• FIBERS demonstrates superior performance in stratifying kidney graft failure risk compared to traditional methods.
• This novel approach offers a more precise immunogenetic basis for risk assessment in kidney transplantation.
• Further research into specific HLA AA-MMs can refine transplant outcomes.