Fig. 3

The influence of pre-sort abundances on existing IgA-Seq scores can be overcome using probability-based approaches. a A toy example where two samples have two species with the same IgA binding profiles but different starting abundances. b The distribution of arbitrary IgA binding values used for the ten species in the IgA-Seq simulations, dashed lines indicate the IgA⁺ and IgA⁻ fraction thresholds. Species coloured as in (c). c The species abundances in the pre-sorted, IgA⁺ and IgA⁻ samples for the 30 samples simulated. d The size of the IgA⁺ and IgA⁻ fractions in the simulated samples as a percentage of their total bacteria. e Comparison of variance in IgA binding estimates across samples using each scoring approach. Coefficient of variance was calculated for each species individually. **p < 0.01, ****p < 0.0001 in Mann-Whitney tests after FDR correction. Boxes represent the median and interquartile range (IQR) and the whiskers the largest and smallest values within one and half IQR of the upper and lower IQR limits. f Pearson and Spearman correlations between the true mean IgA binding scores used to simulate data and the scores estimated by each of the different indices