Population-specific Graph Genome Framework for Identifying Genetic Determinants of Neurological Disorders and Chronic Kidney Disease in the Saudi Arabian Cohort

Abstract Background: Conventional linear reference genomes, such as GRCh38, often fail to capture population-specific genetic variation, leading to alignment biases and reduced variant detection accuracy. This limitation has a significant effect on the underrepresented populations such as the Saudi Arabian cohort where unique variants may play an important role in the pathogenesis of neurological disorders and chronic kidney disease. Methods: We developed a population-specific weighted sequence graph genome incorporating variants found in Saudi. The graph uses minigraph-based pangenome construction, mapping edge weights to allele frequency and predicted pathogenicity. A graph convolutional network (GCN) was implemented for variant prioritisation, with hybrid alignment-scoring function controlling read mapping. The pipeline was designed to work with existing gnomAD tools like GATK to ensure compatibility with clinical workflow. Results: Analysis of whole-genome sequencing data from 1250 individuals showed a 12.7% improvement in variant calling sensitivity and a 37.2% reduction in reference bias compared to linear reference based approaches. The framework uncovered population-enriched variants, such as a 23 bp deletion in CUBN associated with albuminuria and a missense variant in DEPDC5 related to familial epilepsy. The GCN accurately differentiated loci linked to phenotype, with high concordance. Conclusions: This framework improves the detection and prioritisation of variants in less-studied populations. It is a scalable, clinically compatible solution to develop precision medicine that helps to overcome limitations of global reference genome.