Principal component analysis with linear and quadratic discriminant analysis for identification of cancer samples based on mass spectrometry

Abstract

Mass spectrometry (MS) is a powerful technique that can provide the biochemical signature of a wide range of biological materials such as cells and biofluids. However, MS data usually has a large range of variables which may lead to difficulties in discriminatory analysis and may require high computational cost. In this paper, principal component analysis with linear discriminant analysis (PCA-LDA) and quadratic discriminant analysis (PCA-QDA) were applied for discrimination between healthy control and cancer samples (ovarian and prostate cancer) based on MS data sets. In addition, an identification of prostate cancer subtypes was performed. The results obtained herein were very satisfactory, especially for PCA-QDA. Selectivity and specificity were found in a range of 90-100%, being equal or superior to support vector machines (SVM)-based algorithms. These techniques provided reliable identification of cancer samples which may lead to fast and less-invasive clinical procedures.