Prediction of Systemic Lupus Erythematosus from Genotype and Gene-Expression Data: Batch Correction and Leakage-Free Evaluation
Year: 2026
Type: preprint
Abstract: Systemic Lupus Erythematosus (SLE) is a complex autoimmune disease with a strong genetic component and highly variable clinical features. Because of this heterogeneity, predicting disease status remains challenging using any single marker alone. In this study we explore two com-plementary data sources, (1) genotypes at 30 candidate SNPs in a South Indian case–control cohort ( n = 425) and (2) bulk expression of 13 genes from the same pathways, curated from six GEO series ( n = 642 SLE or control samples from 483 individuals), to investigate whether they can be used to distinguish SLE cases from healthy controls. The expression data were placed on a common log 2 scale and corrected for batch effects with mean-only ComBat. To avoid data leakage, ComBat, gene filtering, scaling and model fitting were re-estimated inside every train-ing fold, and all samples from the same individual were kept in the same fold. Performance was summarised with ROC-AUC, precision–recall AUC (PR-AUC) and the Matthews correlation coefficient (MCC) over 10 × 5-fold cross-validation, the models were interpreted with out-of-fold SHAP values, and the whole pipeline was checked with label permutations. For the SNP data, a dominant encoding with age and sex gave ROC-AUC 0.699 ± 0.052, PR-AUC 0.668 ± 0.063 and MCC 0.309 ± 0.101, but this depended on four SNPs that deviate from Hardy–Weinberg equilibrium in controls; without them the SNPs added little to age and sex (ROC-AUC 0.656). The strongest single association, at MTRR A66G, came from an excess of heterozygotes among controls, a pattern typical of genotyping error, whereas homozygous CYP1A1 M2 and COMT genotypes and CYP1A1 M4 carriers were enriched in cases, pointing to xenobiotic metabolism. The leakage-free expression classifiers reached ROC-AUC 0.837 ± 0.033 (L1-penalised logistic regression) and 0.915 ± 0.021 (XGBoost), and shuffled labels gave chance-level results, confirming that the pipeline does not leak. However, a model given only the study label reached ROC-AUC 0.857, because disease status is strongly confounded with study. Within individual studies the ROC-AUC was 0.743 (L1-LR) and 0.835 (XGBoost), and it fell to 0.46–0.74 when whole studies were held out. Lower MTR expression was the most important and most consistent expression feature, whereas two other highly ranked genes, CYP1A1 and GSTT1 , carried little signal within studies. For pooled multi-study data, leakage-free cross-validation alone does not guarantee a reliable estimate; batch–label confounding and genotype quality have to be checked directly.
Institution: Indian Institute of Technology Madras
Language: English
Subfield: Rheumatology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7223458612