Fig. 4

End-to-end analysis of growth-based MAVE datasets with satmut_utils and DiMSum. SUMO1: Homo sapiens small ubiquitin-like modifier 1. HA: influenza A/WSN/1933(H1N1) hemagglutinin. In all panels, all six amplicons in the HA dataset [30] were analyzed, and only one of the three targeted amplicons in the SUMO1 dataset [4] was analyzed for simplicity (Methods). For A–F, only “single” variants are plotted (those altering only one codon/amino acid). For C, D, H, and I, the Pearson correlation coefficient for DiMSum fitness estimates is indicated. In E and F, outliers are greater than 1.5 * interquartile range. A Variant counts by software method. A pseudocount of 1 was added. B Variant call overlap by software. C SUMO1, and D HA fitness estimates by software. E SUMO1, and F HA fitness estimate distributions by variant type. Asterisks indicate significant two-sided Wilcoxon rank-sum tests at a threshold of < 1 × 10⁻¹⁰. G Biological replicate correlation. Shown is correlation between fitness estimates, input counts, and output counts. The SUMO1 dataset contained two replicates. For HA dataset triplicates, the mean correlation + / − the standard deviation is shown. H DiMSum, and I satmut_utils replicate correlation for SUMO1. Scatterplots derive from DiMSum reports which aggregate counts for visualization