In a first for machine-learning algorithms, a new piece of software developed at Caltech can predict behavior of bacteria by reading the content of a gene. The breakthrough could have significant implications for our understanding of bacterial biochemistry and for the development of new medications.