Unsupervised learning methods and similarity analysis in chemoinformatics

Abstract

In this chapter, we present an overview of various chemometric methods, appropriate for analyzing and interpreting data from social media, industry, academia, medicine, and other sources. We discuss unsupervised machine-learning techniques used for grouping (hierarchical cluster analysis, k-means) and exploring (principal component analysis, self-organizing Kohonen maps) all types of data, both quantitative and qualitative. For each method described in this chapter, we explain the basic concepts, provide a rudimentary algorithm, and present practical applications. All the examples are based on a set of molecular descriptors calculated for a selected group of persistent organic pollutants (POPs).