RT Journal Article SR Electronic T1 Phenotype prediction in an Escherichia coli strain panel JF bioRxiv FD Cold Spring Harbor Laboratory SP 141879 DO 10.1101/141879 A1 Marco Galardin A1 Alexandra Koumoutsi A1 Lucia Herrera-Dominguez A1 Juan Antonio Cordero Varela A1 Anja Telzerow A1 Omar Wagih A1 Morgane Wartel A1 Olivier Clermont A1 Erick Denamur A1 Athanasios Typas A1 Pedro Beltrao YR 2017 UL http://biorxiv.org/content/early/2017/05/24/141879.abstract AB Understanding how genetic variation contributes to phenotypic differences is a fundamental question in biology. Here, we set to predict fitness defects of an individual using mechanistic models of the impact of genetic variants combined with prior knowledge of gene function. We assembled a diverse panel of 696 Escherichia coli strains for which we obtained genomes and measured growth phenotypes in 214 conditions. We integrated variant effect predictors to derive gene-level probabilities of loss of function for every gene across strains. We combined these probabilities with information on conditional gene essentiality in the reference K-12 strain to predict the strains’ growth defects, providing significant predictions for up to 38% of tested conditions. The putative causal variants were validated in complementation assays highlighting commonly perturbed pathways in evolution for the emergence of growth phenotypes. Altogether, our work illustrates the power of integrating high-throughput gene function assays to predict the phenotypes of individuals.HighlightsAssembled a reference panel of E. coli strainsGenotyped and high-throughput phenotyped the E. coli reference strain panelReliably predicted the impact of genetic variants in up to 38% of tested conditionsHighlighted common genetic pathways for the emergence of deleterious phenotypes