Clustering sequences into clonal lineages

Sequences with similar CDR3 are possibly progenies from the same naive B cell and can be grouped into a clonal lineage. To detect the lineage structure for the antibody repertoire, we performed single linkage clustering, using a re-parameterization of the method described previously^{5, 6, 44}, accounting for the larger size of the CDR3 and junction in humans as compared to zebrafish. RNA sequences with the same V and J allele assignments, the same junction length, and whose junction regions differed by no more than 10% on the nucleotide level were grouped together into a lineage. This is equivalent to a biological clone that underwent clonal expansion. To test the robustness of this threshold, we also tried the threshold of 95% similarity for CDR3 region⁴⁴, and it did not change the overall position of each lineage in the diversity-size plot, nor did it change the linear fits of lineage distribution generated from these two threshold (Supplementary Fig. ¹²). Lineage diversity is the number of unique RNA molecules within the lineage, and lineage size is the total number of RNA molecules within the lineage. Samples with 5,000,000 PBMCs were subsampled to 120,000 RNA molecules. Samples with fewer PBMCs were subsampled to proportionally fewer RNA molecules according to the PBMC number.