Fig. 3: Associations of coding mutations and CNAs with disease progression.
a, Three cohorts used for studying the presence of variants during disease evolution. Unpaired samples are taken from different patients; cohort (1) were samples from treatment-naïve patients and R/R patients; cohort (2) were paired samples of CLL and RS phase of the same patient; cohort (3) were paired samples taken at two different timepoints before treatment and at relapse. b, Distribution of cancer cell fractions in the three cohorts studied for selected genes. For cohort (1), figures are not shown if no R/R sample carried a mutation. Other genes are presented in Extended Data Fig. 4e,f. Boxplots showing results for unpaired samples and connected datapoints show results for paired samples (corresponding variants are connected by a dotted line). An asterisk indicates a candidate driver. c, Genomic features linked to patients’ PFS (left panel) and OS (right panel). Hazard ratio and FDR of each genomic feature tested against PFS using a Cox proportional-hazards model on the subset of patients for which clinical outcomes data were available (n = 243). Adjusted P values (FDR) are shown in different colors. (See Supplementary Table 14 for the full detailed list of genomic features tested and Supplementary Table 15 and 16 for full results of the statistical tests). d–f, candidate driver IRF2BP2 was recurrently affected by CN losses (d) and SNVs/indels, especially truncating ones (e), and was associated with increased CCF in variants for more advanced disease stage (f). Coloured rectangle in (e) represents protein domains. g–j, candidate driver SMCHD1 was recurrently affected by CN losses (g) SNVs/indels, especially truncating ones (h), presented evidence of increased CCFs in more advanced CLL in cohort (2) (no data available in R/R of cohort (1)) (1), and associated with more adverse overall survival as shown in the Kaplan–Meier plot where shaded areas show the 95% confidence intervals and P values were derived from a log-rank test (j). Coloured rectangle in e represent protein domains. Boxplots show the minimum and maximum values and interquartile range and each individual variant is represented with an individual datapoint.
