Fig. 2: Overview of somatic RT calls in TCGA.

a Count of calls in call set across all tumor types, annotated as either canonical or inversion-containing, as an absolute count (bottom axis) and percent of all calls (top axis). Gray bars indicate calls for which both callers share the same annotation. Blue and orange bars are additional calls annotated in each group by only xTea or only TotalReCall, respectively. b Percentage of insertion events containing inversions by indication for indications with ≥500 insertions total. Error bars and the bar height represent the 99% confidence interval and the estimate of the mean using the Clopper–Pearson method. c Estimated length of the inserted L1 within the canonical somatic RT calls. Length estimates taken from TotalReCall. Gray, truncated insertions. Blue, full-length insertions. Left axis, absolute count of calls. Right axis, percent of total canonical somatic RT calls within each bin. Length does not include transduction region, if one is present. d Somatic RT count per sample grouped by tumor type. Center line indicates median. Box indicates interquartile range. Points more than 1.5× IQR away from the box are shown as individual outliers. Tumor types are sorted in descending order by median somatic RT burden. Mean and median RT count per sample by tumor type listed in boxes above the tumor type name. Red bars indicate tumor types that are significantly enriched in RT (see Ns for each significantly enriched tumor type in figure), compared to a background of all samples from tumor types indicated by gray bars (N = 2265). All significant tumor types had p < 10⁻¹⁰ from a two-sided Mann–Whitney U test (effect sizes > 0.57). Cervical cancer samples also had a significant p-value (p = 4 × 10⁻⁸) from the two-sided Mann–Whitney U test (effect size = 0.94) but are not considered significant here due to the small sample size (N = 3 tumor-normal pairs).