Identification of Alzheimer’s disease using gene expression data - Shehan Shaman Perera

Abstract:
Treatment of Alzheimer’s disease is significantly hampered by the lack of reliable biomarkers that can detect disease traits at an early stage. Microarray gene expression data can be effectively used for such biomarkers identification. Widely used methods in extracting information from this data include ranking the genes according to their differential expressions. Often, features selected on this criterion have redundancy. To address this issue, we propose a new machine learning-based pipeline that extracts a subset of genes from principal component analysis, feature importance scores of random forest, extra tree classifier, and correlation values among genes. The proposed pipeline enabled extraction of 14 potential biomarker genes with 97% accuracy from 161 gene expression profiles. Compared to other methods, this approach has the advantage of minimized redundancy and higher accuracy.