Estimated number of beneficial mutations within each MIC category.

ML estimates and their 95% confidence intervals are based on the observed number of unique mutations relative to the sample size. Two corrections are introduced. (1) Some mutations are identified in multiple MIC categories and are subtracted from the category in which they are most rare (see Table S2 for details). (2) The error-prone polymerase displays mutation bias (see Table S1). Rare transversions are underrepresented and we extrapolated the estimates for the common transitions and transversions.