Detection and analysis of ancient segmental duplications in mammalian genomes

Abstract

Although segmental duplications (SDs) represent hotbeds for genomic rearrangements and emergence of new genes, there are still no easy-to-use tools for identifying SDs. Moreover, while most previous studies focused on recently emerged SDs, detection of ancient SDs remains an open problem. We developed an SDquest algorithm for SD finding and applied it to analyzing SDs in human, gorilla, and mouse genomes. Our results demonstrate that previous studies missed many SDs in these genomes and show that SDs account for at least 6.05% of the human genome (version hg19), a 17% increase as compared to the previous estimate. Moreover, SDquest classified 6.42% of the latest GRCh38 version of the human genome as SDs, a large increase as compared to previous studies. We thus propose to re-evaluate evolution of SDs based on their accurate representation across multiple genomes. Toward this goal, we analyzed the complex mosaic structure of SDs and decomposed mosaic SDs into elementary SDs, a prerequisite for follow-up evolutionary analysis. We also introduced the concept of the breakpoint graph of mosaic SDs that revealed SD hotspots and suggested that some SDs may have originated from circular extrachromosomal DNA (ecDNA), not unlike ecDNA that contributes to accelerated evolution in cancer.

mosaicFlye: Resolving long mosaic repeats using long error-prone reads

“…Since there are no sources and sinks in the complex subgraph (otherwise, a source or a sink vertex would be semi-complex), each connected component of the complex subgraph contains a directed cycle. Such connected components represent the most complex mosaic repeats (referred to as cyclorepeats ) such as ultralong tandem repeats or cyclic segmental duplications analyzed in Pu et al, 2018. With the exception of the recently proposed algorithm for assembling centromeres , existing repeat resolution tools (including mosaicFlye) are unable to resolve cyclorepeats.…”

“…We will first consider the case when the set Fitting ( Read, Genome ) includes only two alignments A 1 and A 2 . Repeats in a genome accumulate mutations and result in divergent repeat copies, e.g., most segmental duplications in the human genome diverge by more than 1% (Pu et al, 2018). As a result, the correct alignment of a read (to a repeat copy that it originated from) typically has a larger percent identity than the alignment of the same read to an incorrect repeat copy.…”

“…Repeats in a genome accumulate mutations and result in divergent repeat copies, e.g., most segmental duplications in the human genome diverge by more than 1% (Pu et al, 2018). Vollger et al, 2019 described how to use the variations between various repeat copies for reconstructing all copies of a divergent repeat.…”

“…First, the number of copies (edge multiplicity in the assembly graph) is often unknown and may vary along a repeat. Second, unlike the haplotype assembly, where all haplomes align to a consensus sequence, many repeats have complex mosaic structure (Pevzner et al, 2004, Jiang et al, 2007, Pu et al, 2018 that prevents utilization of a single consensus sequence as a template for aligning all copies of a repeat. Such mosaic repeats are also common in cancer genomes, making it difficult to analyze duplications that represent the hallmarks of many cancers (Nattestad, 2018).…”

“…However, mosaic repeats consist of several smaller sub-repeats that appear with varying multiplicities and in different combinations within various copies of a mosaic repeat ( Figure 1). Such mosaic repeats are common, e.g., most segmental duplications in the human genome (Pu et al, 2018) and many repeats in bacterial genomes (Pevzner et al, 2004) represent mosaic repeats.…”