Fig. 4: Transcriptional strand asymmetry at insertions and deletions.

a Transcriptional strand asymmetry of insertions and deletions at polyT tracts. Error bars represent standard deviation from bootstrapping with replacement. Both insertions and deletions displayed a strand asymmetry bias towards the non-template strand for polyT tracts across cancer types (binomial test with Bonferroni correction, p value < 0.001 for insertions and p value < 0.05 for deletions). b Transcriptional strand asymmetry occurring at polyT tracts according to level of gene expression for insertions. Mann–Whitney U with Bonferroni correction, p value < 0.001 when comparing low and high expression gene sets across all cancer types except skin, ovarian and lymphoid cancers (p value < 0.05) and CNS (p value > 0.05). c Transcriptional strand asymmetry occurring at polyT tracts according to level of gene expression for deletions. Mann–Whitney U with Bonferroni correction, p value < 0.001 when comparing low and high expression gene sets for skin and p value < 0.05 for stomach and pancreatic cancers. d Hierarchical clustering displaying transcriptional strand asymmetries for indels overlapping dinucleotide motifs. Dinucleotide repeat tracts of up to five repeated units are displayed. Purple represents asymmetry towards the non-template strand, whereas orange represents asymmetry towards the template strand. In the dendrogram of cancers, biliary, uterus, colorectal and stomach cancers are more distant from the other cancers, and contain MSI samples, while lung cancers are also separable from other cancer types, further reinforcing our observations regarding the DNA damage and repair processes that contribute to the observed asymmetries. Across cancer types a non-template strand asymmetry preference was observed for TG, TC and CT motifs (binomial test with Bonferroni correction, p value < 0.001) and for GT motifs (binomial test with Bonferroni correction, p value < 0.05) and a template strand asymmetry for CA, GA and AG motifs (binomial test with Bonferroni correction, p value < 0.001) and for AC motifs (binomial test with Bonferroni correction, p value < 0.05).