High-throughput laboratory evolution with machine learning reveals constraints for drug resistance evolution

The understanding of evolution is crucial to tackle the problem of antibiotic resistance which is a growing health concern. Although the lack of sufficient data has long hindered the mechanism of evolution, laboratory evolution experiments equipped with high-throughput sequencing/phenotyping are now gradually changing this situation. The emerging data from recent laboratory evolution experiments have revealed repeatable features in evolutionary processes, suggesting the existence of constraints on evolutionary outcomes [1,2]. Despite its importance for understanding evolution, however, we still lack a systematic investigation for evolutionary constraints. In this seminar, I would like to talk about two projects on the investigation of evolutionary constraints using data acquired from laboratory evolution of Escherichia coli. In the first half, I will explain how to extract an effective latent space for probing constraints in resistance evolution using gene expression data. We will further discuss what kind of structure exists in this space [3]. In the latter half, I will talk about our recent study on how to construct a predictive model for evolution using the information of evolutionary constraints.