A simple clustering approach for pathogenic strain identification based on local and global amino acid compositional signatures from enomic sequences: the Escherichia genus case

Cluster analysis offers a suite of powerful unsupervised methods, commonly used as exploratory data analysis tools. Such tools can be proven especially useful when we face the situation of analyzing large data sets and want to get an intuitive insight at subtle correlations between instances of the data. In this work, we demonstrate that simple hierarchical clustering approaches (based on compositional features extracted from the amino acid sequences encoded in the complete genomic sequences of 25 species/strains belonging to the proteobacterial genus Escherichia) can be used to accurately discriminate between pathogenic and nonpathogenic strains of those bacteria.