MimicryDB-Auto: Structural Validation Reveals the Inadequacy of Sequence-Based Molecular Mimicry Screening in Autoimmunity

Read full text Summarise with VIRAT
Authors: Minza Ilahi
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Computational identification of molecular mimicry—a key mechanism linking infection to autoimmune disease—has historically relied on sequence-based metrics (BLAST, percentage identity, BLOSUM matrices) under the untested assumption that sequence similarity predicts structural similarity at the MHC-presented peptide level. This study presents MimicryDB-Auto, the first curated, labelled multi-pathogen dataset integrating MHC epitope prediction, sequence alignment, and atomic structural validation for 399 pathogen-host peptide pairs spanning 32 organisms associated with autoimmune rheumatic diseases and Guillain-Barré syndrome. Structural validation using TM-align reveals that within a high-confidence sequence-similar pool (≥50% identity, ≥90% coverage), sequence identity explains at most 1.6% of variance in structural RMSD (r = −0.127, R² = 0.016)—a relationship of no practical predictive utility. Cross-pairing validation demonstrates that 91.9% of randomly assigned sequence-similar 9-aa pairs achieve structural equivalence (RMSD < 2.0 Å), and 99.2% of structurally equivalent cross-pairs have zero detectable sequence similarity, quantifying the scope of sequence-dissimilar structural mimicry invisible to conventional screening. A Random Forest classifier trained on sequence and immunological features achieves near-perfect classification when separating sequence-rich from sequence-poor pairs (AUC = 0.958) but performance drops substantially (AUC = 0.841) on the biologically relevant task of distinguishing strong from weak mimics within the sequence-similar pool, confirming that sequence features cannot substitute for structural validation. These findings challenge the foundational assumption of sequence-based mimicry screening and establish structure-first validation as an essential standard for computational autoimmunity research. MimicryDB-Auto and the complete reproducible pipeline are publicly available at https://github.com/minbaku/molecular-mimicry-RA-pipeline.

Fields of Study & Subjects

vaccines and immunoinformatics approaches Protein Structure and Dynamics Computational Drug Discovery Methods

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
27 March 2026
Language
EN
Document type
Preprint