Feature selection and validation for mass spectral data in proteomics

Description/Abstract

In the area of proteomics, one of the applications is to detect a given type of disease on the basis of patient blood samples. A well-recognized challenge in classification for this type of problem is that there are thousands of features, but only a limited number of samples available. Thus, feature selection becomes an essential procedure to prevent over-fitting before any classification model can be built. We discuss various ways in selecting the key features from the feature space, with respect to the performance of the classification model, as well as the biological implication