CRISPR: a computational framework for personalized type 2 diabetes prevention using GWAS-based polygenic risk

Type 2 Diabetes Mellitus (T2D) is a complex polygenic metabolic disease that is predisposed by highly complex interactions between gene-controlled predisposition and physiological factors. Although the machine learning (ML) methods illustrated high clinical risks prediction capabilities, the combination with the genetic health risks (inherited) and genome-editing design has not been utilized. This paper introduces a complete in silico computational system comprising Polygenic Risk Scoring (PRS) based on Genome-Wide Associations, ML-based risk classification based on T2D and AIguided CRISPR guide RNA (gRNA) prioritization as a methodological proof-of-concept. Polygenic risk is calculated with a transparent GWAS weighted formulation built upon the results of summary statistics and clinical features to supervised ML modeling. SHAP analysis provides model interpretability. Highimpact GWAS loci are computationally ranked to be analyzed in downstream CRISPR analysis, where candidate gRNAs are ranked with forecasted efficiency and off-target risk. The model of the Random Forest showed that the generalization was stable, and the test AUC was 0.8302, with T C F 7 L 2 becoming the domineering locus in the region that was examined. All the results are strictly computational and do not presuppose clinical readiness, serve solely as a reproducible framework in the future research that aims at exploring the translation.