Extended Data Fig. 2: Comparison of computational correction method with UMIs for PCR artefacts and sequencing reproducibility on different occasions.

Germ-free mice were either orally or systemically primed three times every other day by intragastric (10¹⁰ CFU) or intravenous (10⁸ CFU) exposure to E. coli HA107 and compared with germ-free control mice. Immunoglobulin repertoire sequencing at 21 days for IgA in MLN and IgG2b in spleen was performed in parallel from the same RNA samples with two different primers containing different UMIs or primers without UMIs. a, b, MDS plot of MLN IgA (a) or spleen IgG2b (b) repertoires showing Euclidean distance between points representing the similarity between results whether computational correction (filled symbols) or UMI correction (open and cross symbols for the different UMIs) was used. c, Heat map based on CDR3 sequence identity reflecting similarity between UMI-corrected and computational-corrected IgG2b repertoires. d, Identity matrix based on CDR3 sequences of technical replicates of the same biological sample that had been used on two separate occasions for the entire pipeline of cDNA preparation, IgA amplicon PCR, library preparation and MiSeq sequencing. The distinctions between replicate mice and technical repeats are shown in the diagram.