Autoencoder with salp optimisation technique for exploring SNP-SNP interactions in Alzheimer's disease

Genetic association research aims to identify genetic variations linked to specific disease states or conditions. Genetic interactions, also known as epistasis, typically involve interactions among numerous single nucleotide polymorphisms (SNPs). Detecting genetic interactions among millions of SNPs in GWAS is challenging. This study presents a two-stage epistasis model called autoencoder based feature selection with Salp optimiser for Epistasis identification (AE-SalpEpi). The autoencoder (AE) approach is designed in the screen phase to pick a subset of features having a high association with Alzheimer's disease. During the selection stage, disease-correlated SNP combinations are chosen using SalpEpi-SO and SalpEpi-MO. The simulated and real Alzheimer's disease dataset are used to quantify the performance of AE-SalpEpi-SO and AE-SalpEpi-MO and also contrasted to state-of-art techniques such as MCASO-Epi and MACOED. Experimental results showed that the suggested methods successfully identified interactions with the APOE gene variation, a known risk factor for Alzheimer's disease.