DNAi: AI based disease risk and genetic insight predictor through CNN, Random Forest

DNAi: AI-Based Disease Risk and Genetic Insight Predictor Through CNN, Random Forest, and BiLSTM DNAi is an open-source hybrid AI framework for classifying disease-associated Single Nucleotide Polymorphisms (SNPs) across three high-burden cardiometabolic conditions — Type 2 Diabetes Mellitus (T2DM), Coronary Artery Disease (CAD), and Hypertension (HTN). The system curates 20 GWAS-validated RSIDs anchored to known molecular pathways and encodes 501 bp flanking sequences (GRCh38) using five nucleotide encoding strategies. A Convolutional Neural Network (CNN) trained on one-hot encoded sequences achieves 89.4% overall accuracy and a ROC-AUC of 0.961, while a Random Forest classifier provides complementary SHAP-based per-variant interpretability. Top attributions identify TCF7L2 (rs7903146), CDKN2A/B (rs10757278), and ACE (rs4646994) as the highest-impact variants for T2DM, CAD, and HTN respectively. The full pipeline is deployed as a RESTful web application (FastAPI backend, TypeScript frontend), enabling browser-based genomic risk prediction without local computational infrastructure. Keywords: DNA sequence classification, SNP, GWAS, Type 2 Diabetes, Coronary Artery Disease, Hypertension, CNN, Random Forest, SHAP, bioinformatics, genomic risk, FastAPI