Theoretical Modelling of Polygenic Inheritance and Disease Risk Prediction Across Admixed Nigerian Populations

Polygenic risk scores (PRS) for complex diseases are typically derived from European-ancestry GWAS, and their application to admixed African populations produces substantially attenuated predictive performance due to trans-ancestry differences in linkage disequilibrium, causal allele frequencies, and population stratification. No theoretical framework specifically addresses PRS portability across the admixed and genetically heterogeneous Nigerian population landscape. This dissertation develops a novel theoretical model for polygenic inheritance and disease risk prediction optimised for admixed Nigerian populations, with empirical validation in hypertension and type 2 diabetes phenotypes. The theoretical contribution integrates local ancestry-informed PRS construction (LA-PRS), Bayesian transfer learning across ancestry-matched GWAS summary statistics, and simulation-based power modelling for multi-ancestry PRS meta-analysis in Nigerian cohorts. Empirical validation employed genome-wide data from 2,000 individuals spanning Hausa-Fulani, Yoruba, Igbo, and admixed participants from all six geopolitical zones. The LA-PRS model significantly outperformed ancestry-blind PRS for both hypertension and T2DM across admixed individuals. Theoretical derivations of optimal ancestry weighting under various admixture architectures were validated by simulation. This dissertation advances a generalisable theoretical framework for polygenic risk prediction that accommodates the unique genetic architecture of Nigerian and broader West African admixed populations. Keywords: polygenic risk score, theoretical model, admixed populations, local ancestry, Nigeria