Identifying the candidate genes using co-expression, GO, and machine learning techniques for Alzheimer's disease

Alzheimer's disease is a neurological disorder that affects an individual's memory, motor functions, behaviour, and thought process. It has been observed that the hippocampus is the first region that gets affected by Alzheimer's. Hence, a study of the hippocampus region can identify genes responsible for the occurrence of the early stage of the disease. Most often, t-test and correlation are used to identify significant genes at the initial level. As the genes are differentially expressed, their classification power is generally high. These genes might appear significant, but their degree of specificity towards the disease might be low, leading to misleading interpretations. Similarly, there may be many false correlations between the genes that can affect the identification of relevant genes. This paper introduces a new framework to reduce the false correlations and find the potential biomarkers for the disease. The framework concerned uses the t-test, correlation, Gene Ontology (GO) categories, and machine learning techniques to find potential genes. The proposed framework detects Alzheimer-related genes and achieves more than 95% classification accuracy in every dataset considered. Some of the identified genes which are directly involved in Alzheimer are APP, GRIN2B, and APLP2. The proposed framework also identifies genes like ZNF621, RTF1, DCH1, and ERBB4, which may play an important role in Alzheimer's. Gene set enrichment analysis (GSEA) is also carried out to determine the major GO categories: down-regulated and up-regulated.