Extended Data Fig. 2: Time to recurrence analysis of genomic metrics.

(A) Participants are split between those with >1 mutated gene on the panel (red, n = 29 participants) and those with 1 or 0 (grey, n = 66 participants). This threshold produced the best split of the data in time to recurrence analysis. (B) Participants are split by the median mPGA. Participants with high mPGA (red, n = 53 participants) do not have significantly shorter time to recurrence than those with low mPGA (grey, n = 53 participants). (C) Participants are split by the median max PGA. Participants with high max PGA (red, n = 53 participants) do not have significantly shorter time to recurrence than those with low max PGA (grey, n = 53 participants). (D) Participants are split between those that had a subclonal mutation on the driver gene panel (red, n = 38 participants) and those without (grey, n = 57 participants). KMT2C and KMT2D were excluded from this analysis. (E) Participant with an amplification in MYC and/or FGFR1 (red, n = 5 participants) did not have significantly different time to recurrence than those with an absence of either amplification (grey, n = 101 participants). All p values are calculated using a log-rank test. (F) Participants with strong sidedness (greater clustering of right and left regions across the MEDICC2 tree, lambda > 0.8, red, n = 36 participants) show a significantly shorter time to recurrence (log-rank test) than the remaining participants for which phylogenetic signal analysis was possible (grey, n = 31 participants). (G) Cox proportional hazards (CPH) model of time to recurrence using clinical co-variates, phylosig sidedness classification and Spearman (n = 67 participants). Forest plot shows 95% confidence interval of hazard ratios, and the covariate P values, derived from a Wald test (*P < 0.05, ** P < 0.01, ***P < 0.001). Hazard ratio for Spearman represents the increase in hazard between the 5th and 95th percentile values (within the Sequencing Cohort).

Source data