Repeat-rich regions cause false positive detection of NUMTs - a case study in amphibians using an improved cane toad reference genome

Mitochondrial DNA (mtDNA) has been widely used in genetics research for decades. Contamination from nuclear DNA of mitochondrial origin (NUMT) can confound studies of phylogenetic relationships and mtDNA heteroplasmy. Homology searches with mtDNA are widely used to detect NUMTs in the nuclear genome. Nevertheless, false positive detection of NUMTs is common when handling repeat-rich sequences, whilst fragmented genomes might result in missing true NUMTs. In this study, we investigated different NUMT detection methods and how the quality of the genome assembly affects them. We presented an improved nuclear genome assembly (aRhiMar1.3) of the invasive cane toad (Rhinella marina) with additional long-read Nanopore and 10x linked-read sequencing. The final assembly was 3.47 Gb in length with 91.3% of tetrapod universal single-copy orthologs (n=5,310), indicating the gene-containing regions were well assembled. We used three complementary methods (NUMTFinder, dinumt and PALMER) to study the NUMT landscape of the cane toad genome. All three methods yielded consistent results, showing very few NUMTs in the cane toad genome. Furthermore, we expanded NUMT detection analyses to other amphibians and confirmed a weak relationship between genome size and the number of NUMTs present in the nuclear genome. Amphibians are repeat-rich, and we show that the number of NUMTs found in highly repetitive genomes is prone to inflation when using homology-based detection without filters. Together, this study provides an exemplar of how to robustly identify NUMTs in complex genomes when confounding effects on mtDNA analyses are a concern. SignificanceThis study uses an updated cane toad nuclear genome assembly and multiple NUMT detection methods to confirm a lack of NUMTs that might confound the use of mtDNA as a population genetic marker in the cane toad. We provide an exemplar study for NUMT detection accounting for genome assembly quality and composition, and highlight the risks of using BLASTN-based approaches in highly repetitive nuclear genomes.