Fig. 5: sciCSR predicts BCR isotype distribution in time-course immunization data.
a, Schematic of immunization and follow-up data collection from the Kim et al. dataset. LN, lymph node. b, Isotype distribution of GC B cells for donors 7 (top) and 20 (bottom), across weeks 7, 15 and 29 scBCR-seq data. c,d, sciCSR-predicted CSR fluxes, based on TPT, from and to each isotype. Predictions were made on the basis of week 7 scRNA-seq data to predict scBCR-seq isotype distribution at week 15 (c), or using week 15 scRNA-seq data to predict week 29 scBCR-seq isotypes (d). Left: CSR fluxes expressed as a bubble plot. The colors of bubbles correspond to the amount of flux between the given pair of isotypes predicted by sciCSR, and their sizes correspond to statistical significance of these fluxes compared to predictions generated from randomized transitions (see Main and Methods). Right: the sciCSR-predicted switching events to each isotype, represented by total inward TPT fluxes towards each isotype, are compared against the observed BCR isotype distribution at the subsequent time point. These two distributions were compared using cosine similarity.
