An AI-Driven Bioinformatics Pipeline Combining Quantitative Genetics And Advanced Data Analytics For Neurodegenerative Disease Classification

Abstract

Neurodegenerative diseases, such as Alzheimer disease (AD) and Parkinson disease (PD), are multifactorial biological diseases with heterogeneous genetic structure, which are multifactorial polygenic diseases. Despite the many associations of susceptibility loci, which are genome-wide, mapping the genetic susceptibility into the scale services is a difficult task to carry out. In this paper, an artificial intelligence-based bioinformatics pipeline is described that combines the modelling of quantitative genetics with the capabilities of the state-of-the-art data analytics to classify neurodegenerative diseases in a robust way. transcriptomic profiles and SNP data of genomes were going through stringent quality control, such as filtering minor allele frequency, HardyWeinberg equilibrium, and linkage disequilibrium pruning. To measure genetic predisposition, polygenic risk scores (PRS) and the features derived using quantitative trait loci (QTL) were calculated. Ensemble machine learning and deep neural network models were trained after dimensionality cut and feature selection approach had been employed. A relative analysis to traditional genetic models in terms of logistic regression showed better classification ability with the ensemble structure having a higher discrimination ability. SHAP analyses and enrichment of pathways demonstrated that many synaptic signalling, mitochondrial dysfunction, and neuroinflammatory pathway-related genes were strongly activated and inhibited by the feature attribution analysis. The results show that the combination of quantitative genetics and AI-powered analytics increase the predictive power, maintaining biological interpretation, which can be used to build scalable precision neurogenomics models to predict the early disease risks in a stratified manner.