Model predicts TRB clone abundances.

A: Correlation coefficient of the RBM likelihood with clonal abundances at day 21 post-stimulation in all samples from [16]. B: The correlation coefficient shown is obtained by progressively filtering out the low-abundance sequences from the testing set, as illustrated for the Pt3 NA dataset. C: Scatter plot between clone abundance at day 21 and the RBM likelihood for the full Pt3 NA dataset, showing that the correlation is poor for clones at low counts. The golden dashed line marks the minimal clone abundance that is considered to measure the correlation coefficient reported in A and indicated by the golden dot in B. B-D contain the comparison between the performance of the RBM and SONIA, confirming that capturing sequence correlations (as the RBM does, see also S4 Fig) ensures better prediction of clonal abundance.

(PDF)