Germline Variants Absent from Global Reference Databases in Malawian DLBCL Patients: Implications for Variant Interpretation in African Genomic Medicine

Background: African populations remain underrepresented in global genomic reference databases, complicating germline variant interpretation in clinical and research settings. This concern is particularly relevant in Malawi, where population-matched allele-frequency resources are limited. We assessed germline variants in Malawian patients with diffuse large B-cell lymphoma (DLBCL) that were absent from commonly used global reference databases. Methods: Whole-exome sequencing data from 55 histologically confirmed DLBCL cases recruited at Kamuzu Central Hospital, Lilongwe, were analyzed. Germline variants were called using GATK HaplotypeCaller and annotated with Ensembl VEP v110. Cohort allele frequencies were compared with gnomAD v3.1.2, ExAC v1.0, and the 1000 Genomes Project Phase 3 using African (AFR) and European (EUR) reference subsets. Variants absent from all three databases were designated database-absent variants (DAVs); this designation indicates absence from the queried reference panels and does not establish population restriction. Functional consequence, genomic distribution, and enrichment in immune-related loci were evaluated. Results: We identified 127,456 PASS-filtered germline variants across 55 samples. Of these, 332 variants (0.26%) were absent from gnomAD, ExAC, and 1000 Genomes and were distributed across 289 genes (median, one variant per gene; range, 1-8). Predicted loss-of- function variants comprised 16.9% of DAVs compared with 7.8% of the full call set (odds ratio, 2.41; 95% confidence interval, 1.79-3.24; p < 0.001). Seventy-eight DAVs (23.5%) occurred in HLA class I loci (HLA-A, HLA-B, and HLA-C; Fisher's exact test, p < 0.001). Allele-frequency distributions differed from both AFR and EUR reference populations, with greater divergence from EUR (Kolmogorov-Smirnov D = 0.31) than AFR (D = 0.13). The cohort also showed an increased proportion of low-frequency variants relative to both references. Conclusions: Database-absent germline variants were observed in this Malawian DLBCL cohort, including variants with predicted loss-of-function consequences and variants in HLA class I loci. These findings highlight gaps in the representation of Malawian and Southern African genomes in commonly used reference resources and the resulting potential for uncertainty in variant interpretation. Because this case-only cohort lacked matched population controls and sequencing depth was limited, DAVs should not be interpreted as definitively Malawian-specific or pathogenic. Larger, well-covered studies including matched Malawian and Southern African controls are needed to establish population frequencies and improve equitable genomic interpretation.