It is common in bioinformatics to want to classify things into groups without first knowing what or how many groups there may be. This process is usually known as clustering and is a type of unsupervised machine learning. A common place for this approach is in genomics experiments, particularly RNAseq and related expression technologies. In this recipe, we'll start with a large gene expression dataset of around 150 samples, learn how to estimate how many groups of samples there are, and apply a method to cluster them based on the reduction of dimensionality with Principal Component Analysis (PCA), followed by a k-means cluster.