Structural and Biophysical Profiling of HLA Class II Mismatches Using AlphaFold 3: A Computational Study
- 1St. Panteleimon Hospital, First Lviv Medical Union, 79059 Lviv, Ukraine
- 2Department of Clinical Immunology and Allergology, Danylo Halytsky Lviv National Medical University, 79010 Lviv, Ukraine
- 3Department of Therapy No. 1, Medical Diagnostics, Hematology, and Transfusiology, Faculty of Postgraduate Education, Danylo Halytsky Lviv National Medical University, 79010 Lviv, Ukraine
Abstract
Conventional linear eplet-matching algorithms (HLAMatchmaker, HLAMM) evaluate primary sequence polymorphisms across human leukocyte antigen (HLA) loci without integrating three-dimensional (3D) steric accessibility and local physicochemical constraints. Here, we developed an exploratory computational antigenicity pipeline using AlphaFold 3 to evaluate 10,000 simulated European donor–recipient pairs across pre-computed, full-length HLA class I (HLA-A, -B, -C) and heterodimeric class II (HLA-DR, -DQ, -DP) complexes modeled in a standardized peptide-free conformation. Candidate eplets were sequentially screened using a spatial solvent-accessible surface area threshold (SASA ≥ 10.0 Å2) and an empirical biophysical exclusion gate (5th percentile of absolute net charge |Qnet| ≥ 0.3326 e and hydropathy Htotal ≥ −7.265, calibrated on N = 170 antibody-verified eplets). Across the simulated cohort (n = 10,000), filtration reduced median [Q1; Q3] eplet counts from 19 [12; 26] to 17 [10; 23] for HLA-A, 22 [15; 29] to 18 [12; 24] for HLA-B, 12 [7; 17] to 10 [7; 15] for HLA-C, 15 [10; 19] to 11 [7; 14] for HLA-DR, 12 [6; 17] to 9 [4; 14] for HLA-DQ, and 7 [3; 12] to 6 [2; 9] for HLA-DP. Component ablation demonstrated that 3D steric accessibility (SASA ≥ 10.0 Å2) accounted for the vast majority of eplet exclusions—87.8% in HLA-DR and HLA-DP and 58.2% to 69.1% across HLA-DQ and class I loci—whereas the registry-calibrated biophysical gate acted as a secondary refinement. Within the subset of simulated pairs initially exceeding the arbitrary ≥10-eplet linear cutoff, the proportion dropping below 10 eplets after filtering was low for class I (HLA-A: 4.4%, HLA-B: 8.0%, HLA-C: 13.2%) but substantially higher for class II high-load subsets: HLA-DQ (21.8%), HLA-DR (26.0%), and HLA-DP (33.7% of 3693 baseline high-load pairs). Furthermore, class II loci, particularly HLA-DP, exhibited weakened covariance between heuristic charge and hydropathy indices that persisted after 1:1 count-matched subsampling (ρmatched = −0.268) and partial correlation adjustment (ρpartial = −0.385). This dual-filter pipeline defines a restricted mathematical subset of spatially accessible and biophysically suprathreshold candidate eplets. Because this is a hypothesis-generating in silico framework, multicenter retrospective registry analyses and prospective clinical trials evaluating de novo donor-specific antibody formation, biopsy-proven rejection, and allograft survival are strictly required before clinical translation.
Related
- Structural insights into an aptainhibitor of the SARS-CoV-2 E protein using molecular docking, molecular dynamics, and biophysical analyses
- Biophysical and Computational Evidence for Ibuprofen Binding to and Inhibition of MARK4
- Digital molecular kinetics enabled by large-scale nanopore arrays for multidimensional biophysical profiling
- Structural and biophysical insights into the unique RNase YicC
- Chemical profiling and multitarget computational screening reveal bioactive phytochemical scaffolds from Echiochilon pauciflorum stocks ex wight
BibTeX
@article{bardash2026structural,
title = {{Structural and Biophysical Profiling of HLA Class II Mismatches Using AlphaFold 3: A Computational Study}},
author = {Vladyslav Bardash and Oleh Samchuk and Yaroslav Tolstyak and Hanna Demydenko and Eugen Sklyarov},
journal = {International Journal of Molecular Sciences},
year = {2026},
month = oct,
doi = {10.3390/ijms27219442},
url = {https://www.mdpi.com/1422-0067/27/21/9442},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Structural and Biophysical Profiling of HLA Class II Mismatches Using AlphaFold 3: A Computational Study
AU - Vladyslav Bardash
AU - Oleh Samchuk
AU - Yaroslav Tolstyak
AU - Hanna Demydenko
AU - Eugen Sklyarov
PY - 2026
DA - 2026/10/23
JO - International Journal of Molecular Sciences
DO - 10.3390/ijms27219442
UR - https://www.mdpi.com/1422-0067/27/21/9442
AB - Conventional linear eplet-matching algorithms (HLAMatchmaker, HLAMM) evaluate primary sequence polymorphisms across human leukocyte antigen (HLA) loci without integrating three-dimensional (3D) steric accessibility and local physicochemical constraints. Here, we developed an exploratory computational antigenicity pipeline using AlphaFold 3 to evaluate 10,000 simulated European donor–recipient pairs across pre-computed, full-length HLA class I (HLA-A, -B, -C) and heterodimeric class II (HLA-DR, -DQ, -DP) complexes modeled in a standardized peptide-free conformation. Candidate eplets were sequentially screened using a spatial solvent-accessible surface area threshold (SASA ≥ 10.0 Å2) and an empirical biophysical exclusion gate (5th percentile of absolute net charge |Qnet| ≥ 0.3326 e and hydropathy Htotal ≥ −7.265, calibrated on N = 170 antibody-verified eplets). Across the simulated cohort (n = 10,000), filtration reduced median [Q1; Q3] eplet counts from 19 [12; 26] to 17 [10; 23] for HLA-A, 22 [15; 29] to 18 [12; 24] for HLA-B, 12 [7; 17] to 10 [7; 15] for HLA-C, 15 [10; 19] to 11 [7; 14] for HLA-DR, 12 [6; 17] to 9 [4; 14] for HLA-DQ, and 7 [3; 12] to 6 [2; 9] for HLA-DP. Component ablation demonstrated that 3D steric accessibility (SASA ≥ 10.0 Å2) accounted for the vast majority of eplet exclusions—87.8% in HLA-DR and HLA-DP and 58.2% to 69.1% across HLA-DQ and class I loci—whereas the registry-calibrated biophysical gate acted as a secondary refinement. Within the subset of simulated pairs initially exceeding the arbitrary ≥10-eplet linear cutoff, the proportion dropping below 10 eplets after filtering was low for class I (HLA-A: 4.4%, HLA-B: 8.0%, HLA-C: 13.2%) but substantially higher for class II high-load subsets: HLA-DQ (21.8%), HLA-DR (26.0%), and HLA-DP (33.7% of 3693 baseline high-load pairs). Furthermore, class II loci, particularly HLA-DP, exhibited weakened covariance between heuristic charge and hydropathy indices that persisted after 1:1 count-matched subsampling (ρmatched = −0.268) and partial correlation adjustment (ρpartial = −0.385). This dual-filter pipeline defines a restricted mathematical subset of spatially accessible and biophysically suprathreshold candidate eplets. Because this is a hypothesis-generating in silico framework, multicenter retrospective registry analyses and prospective clinical trials evaluating de novo donor-specific antibody formation, biopsy-proven rejection, and allograft survival are strictly required before clinical translation.
SN - 1422-0067
ER -