Feature Extraction (FE) is a difficult task when the number of features is much larger than the number of samples, although that is a typical situation when biological (big) data is analyzed. This is especially true when FE is stable, independent of the samples considered (stable FE), and is often required. However, the stability of FE has not been considered seriously. In this chapter, the authors demonstrate that Principal Component Analysis (PCA)-based unsupervised FE functions as stable FE. Three bioinformatics applications of PCA-based unsupervised FE—detection of aberrant DNA methylation associated with diseases, biomarker identification using circulating microRNA, and proteomic analysis of bacterial culturing processes—are discussed.
Combination of Principal Component Analysis and Genetic Algorithm for Microbial Biomarker Identification in Obesity
