Extended Data Fig. 7: GWAS study design and results of paired sample analysis.

(a) A schematic indicating the design of the two GWAS analyses conducted in this study with red and blue points indicating within-host polymorphisms. A linear model on the log of the unitig counts per million similar to that commonly used in RNA-seq analyses was used in the paired design while the Pyseer algorithm was used in the standard design. (b) The proportion of resistant isolates following antimicrobial treatment. Those samples within a threshold of 4 weeks (28 days) of a treatment event were classified into the ‘treated’ class. (c) A dot plot indicating the significance and average effect size of unitigs found to be associated with treatment in the analysis of paired samples taken from the same host where a subset have received antimicrobial treatment in between sampling events. Regressions were performed using a linear model with the frequency of the unitig within the host taken as the dependent variable (Methods). The horizontal red lines indicate the expected number of false discoveries (EFD) providing different significance levels to interpret the resulting variant calls.