Machine Learning-Based Selection of Type 2 Diabetes Specific Genes and Microarray Datasets

Type 2 diabetes is a chronic complex disease whose mechanism is still elusive. Gene expression profiling using microarray is routinely used to identify etiologically important genes between diabetics and normal individuals, but this method suffers with some of the limitations such as background noise, low sensitivity, and high dimensionality.. To rectify them, statistical meta-analysis is extensively used by combining multiple datasets. In this study, we highlighted the application of some machine learning algorithms: SVM, LDA, and LASSO for selection of relevant genes and expression datasets after meta-analysis of microarray datasets and also provided biological significance of identified genes in light of published GWAS studies. Association of these selected genes with GWAS could likely point toward the probable link between genome and transcriptome.