Supplementary Figure 3: Grid search on the optimal parameters for clustering B cell CDR3s and the differences among clones.

a–c, Heatmaps show the values of FDR (a), cluster ratio (b), and final score (c) for ranking the clustering results of 500 randomly sampled pairs. The best solution with the highest final score was marked in red boxes. d, CDR3 amino acid length comparison for expanded (n = 2,687,374) and non-expanded (n = 32,631) B cell clones. Expanded clones are complete IgH CDR3s found in any B cell cluster (see Methods), and the remaining complete CDR3s are non-expanded ones. Statistical significance was evaluated by Student’s t-test. The red dots and the numbers show the mean values and the bars extend to 2 s.d. on both sides. The curves in the violin plots show the value distribution. e, Overlap of B cell clonotypes for tumor samples and adjacent normal tissues. We chose matched TCGA tumor and adjacent normal tissue samples with at least 500 clonotypes called by TRUST. As a comparison, we downloaded the top 100 B cell clones of paired normal and tumor samples published by Katoh et al.²². We observed a similar Jaccard index of tumor–normal clonotype overlap between data from TCGA (n = 26) and from Katoh et al.²² (n = 30). The two-sided P value was from Wilcoxon signed-rank test.