Comparative GWAS using global and Indian Reference Panels reveals non-coding drivers of COVID-19 severity and mortality

India remains underrepresented in global genomic studies. We hypothesized that population-specific genetic variants contribute to COVID-19 severity and outcomes, and that the choice of reference panel during imputation impacts Genome-Wide Association Studies (GWAS) resolution. Integrating both global and indigenous reference panels may unravel unique and shared genetic associations that are otherwise missed during standard analyses. In this study, we aimed to perform a comparative GWAS using Indian population-specific (IndiGen) and global (1000 Genomes Project/1KGenomes) reference panels to identify potential genetic loci associated with the COVID-19 differential severity and mortality among the Indian patients. Genomic DNA was extracted and genotyped from the patients who were stratified based on the clinical data capturing COVID-19 symptoms and clinical outcomes. Quality control, liftover, phasing and imputation were performed on the genotypic data. GWAS was performed separately for the severity and mortality phenotypes. Significant loci were functionally annotated using Linkage Disequilibrium (LD) analysis, eQTL mapping, and gene annotation tools. Comparative GWAS with 1KGenomes and IndiGen panels revealed both shared and unique loci. 1KGenomes identified protective variants near MIR4432HG involved in endothelial stability, while IndiGen uncovered risk variants with rs10096505 (SFTPC/BMP1) linked to alveolar collapse and fibrotic remodelling. rs9547631 was common to both panels for mortality, whereas IndiGen-specific risk variants (rs78554880, rs112982286, rs111390553, and rs79900659) were associated with immune dysregulation. Functional annotation of these loci pointed to key biologically plausible links to COVID-19 severity and fatal outcomes. Briefly, the use of an indigenous reference panel improved variant discovery and LD resolution, highlighting that population-specific signals are missed by the generic global datasets. Our findings underscore the importance of inclusive genomic resources for accurate association mapping in the underrepresented populations.