Fig. 2

The workflow of fNUMT and the performance evaluation. a (Left) The preprocess of WGS data for downstream analysis. The WGS clean data were first aligned to rCRS. The mtDNA-mapped reads were then aligned to the reference containing nDNA and rCRS to remove the candidate ref-NUMTs covered by reads that were repetitively mapped to autosome and rCRS. The ref-NUMT free data were subsequently input to fNUMT for detecting non-ref NUTMs and NUMT-FPs. (Right) fNUMT searches the candidate junction reads with one end mapped to nDNA and the other to mtDNA. Then cluster such soft-clipped reads according to the mapping coordination and direction to locate the breakpoints of the inserted mtDNA segments and the insertion site on nDNA. Next, fNUMT assembly all the reads mapped to the candidate mtDNA segments and realign the contig to mtDNA to confirm the occurrence of the non-ref NUTMs and detect the mismatches that were the potential non-ref NUMT-FPs. b The number of non-ref NUMTs detected by NUMTs-detection, dinumt, and fNUMT in the four samples. c, d An illustration of the confirmation of the non-ref NUMTs by long-read sequencing data, taking the non-ref NUMT chrM:59–16089-chr11:49883569 as an example. c The IGV plot of short-read data near chr11:49883569 where the soft-clipped reads that could also map to mtDNA (colored in blue) were observed, meaning the presence of insertion sequences originated from mtDNA near chr11:49883569. d The IGV plot of long-read data near chr11:49883569 where the insertion of sequences with inferred sizes of ~ 530 bp (red box) were observed. e The distance of the breakpoints estimated by NUMTs detection, dinumt an fNUMT over those inferred by long-read data. The p-value was measured by the Wilcox test