Previously there has been no pattern shown for which two genes are more likely to form fusions. Our hypothesis was that there is a correlation between genome structure and gene fusion formation. By looking at population-based TCGA fusion data and genome-wide RNA-DNA data, we observed a co-localization of RNA-DNA interaction and fusion pairs. 5 out of top 10 RNA-DNA interactions colocalize with fusion. Experimentally, 37 out of 42 fusions detected in validation cohort of cancer cells correspond to RNA-DNA interaction hotspots in the normal cells.
The DNA sides of RNA-DNA interaction reads were enriched on DNA loop anchors: 609,364 DNA sides co-localized with anchors while shuffled DNA segments contained 476,889 RNA-DNA reads (Permutation test p-value < 2.2e-16).