Fig. 1

Repertoire subsampling accurately controls for variance inflation. A simulated sequencing dataset was generated by drawing 30 replicate samples from a single pool containing 50 genes of varying prevalence. For each replicate, the number of sequences was chosen randomly, and the total count varied between 3000 and 12,000. a The frequency of each gene was tallied, and the euclidean distance between each pair of replicates was calculated. b Each repertoire was subsampled to the size of the smallest repertoire (n = 3216), and euclidean distance was calculated based on normalized gene frequency in the subsampled dataset. The distance measurement was then averaged across multiple subsampling steps. All distance metrics are compared against the original repertoire size for the smaller repertoire