Factoring and Clustering High Content Data

This chapter reviews several unsupervised learning techniques that have a direct application in HCS assays. Factor analysis is an approach that is used widely in the social sciences, where there is a general problem trying to understand deep personal characteristics through observing surface behaviors. Hierarchical clustering assumes that all samples are related to each other, some are closely related and are therefore members of a cluster, similar clusters are themselves clustered, and all clusters are ultimately part of a single cluster. Principal components analysis (PCA) is a method that reduces dimensionality (the number of features), and is resistant to redundancy; highly redundant features would be compressed into a single component. Unsupervised learning brings powerful data analysis tools that were developed for a range of disciplines and have gained a lot of attention for being part of the “Big Data” movement.