Quantifying microbial fitness in high-throughput experiments

Few concepts are as central to evolution as is fitness, and yet the quantification of fitness is often ambiguous. In particular, high-throughput experiments to measure mutant fitness in microbes are increasingly common but vary widely in their definitions of fitness, which makes their results difficult to compare. What are the consequences of these different fitness statistics, and is there a best way to quantify fitness in a given context? Here we systematize the set of possible fitness statistics according to the following three choices: 1) the encoding of relative abundance (e.g., transforming by a log or logit function), 2) the time scale over which to measure the change in relative abundance, and 3) the choice of a reference subpopulation for calculating relative fitness in bulk competition experiments, such as those using DNA-barcoded mutants. We show that these different choices can lead to significantly different interpretations of mutant fitness, affecting the magnitude of fitness effects, the apparent presence of epistasis, and even the fitness ranking across mutants. This can confound predictions for evolutionary dynamics and gene functions. Altogether our results demonstrate the importance of consistent fitness definitions for reproducible results across experiments.