Identification of correlation between Alzheimer's disease and Type 2 diabetes using nonnegative matrix factorization

Abstract

Alzheimer’s disease (AD) is a complex and heterogeneous disease that can be affected by various genetic factors. The cause of AD is not yet known, and there is no treatment to cure this disease; however, its progression can be slowed. AD is a brain-specific type of diabetes called type 3 diabetes. Several studies have shown that people with type 2 diabetes (T2D) have a higher risk of developing AD. Therefore, it is important to identify subtypes of AD to provide personalized and suitable treatments. In this study, we describe a new approach to identify the correlation between AD and T2D. First, subtypes of AD and T2D are generated using a nonnegative matrix factorization (NMF), which generates clusters containing subsets of genes and samples. Then, candidate genes with the same regulation directions in both diseases are extracted by comparing the cluster pairs generated by AD and T2D. To generate clusters using NMF, we propose a method to select genes that have relative differences between disease patients. Here gene selection criteria based on a nonparametric test are used to select only genes with significant differences between clusters of disease patients. Then, distinct subtypes (genes and samples in the clusters) are generated for AD and T2D. ii Next, we performed a pathway enrichment analysis for the genes in each disease cluster. We found an AD cluster that is related to T2D from the AD samples and a T2D cluster that is related to AD from the T2D samples. In these two clusters, we identified common genes, including 262 relatively up-regulated genes and 67 relatively down-regulated genes. We propose these 329 common genes as candidate genes that are related to both AD and T2D. Further, we identified these 329 candidate genes as playing a role in the common pathological features of AD and T2D. In addition, when we classified T2D patients and normal controls from an independent dataset using these 329 candidate genes and a neural network classifier, we obtained an area under the curve (AUC) value between the true positives and false positive rates of 0.874, which is a significantly high AUC value compared with randomly selected genes (which resulted in an AUC of 0.823) with a p-value of 3.90 × 10−188.