Dot2dot: Accurate Whole-Genome Tandem Repeats Discovery

Abstract

AbstractThe advent of sequencing technologies and the consequent computational analysis of genomes has confirmed the evidence that DNA sequences contain a relevant amount of repetitions. A particularly important category of repeating sequences is that of tandem repeats (TRs). TRs are short, almost identical sequences that lie adjacent to each other. The abundance of TRs in eukaryotic genomes has suggested that they play a role in many cellular processes and, indeed, are also involved in the onset and progress of several genetic disorders.Building upon the idea that similar sequences can be easily displayed using graphical methods, we formalized the structure that TRs induce in dot plot matrices where a sequence is compared with itself. We further observed that a compact representation of these matrices can be built and searched in linear time in the size of the input sequence. Exploiting this observation, we developed an algorithm fast enough to be suitable for whole-genome discovery of tandem repeats.We compared our algorithm with seven state of the art methods using as a gold standard five collections of tandem repeats: pathology-linked, forensic, for population analysis, genealogic-oriented, and variable TRs in regulatory regions. In addition, we run our algorithm on seven reference genomes to test the suitability of our approach for whole-genome analysis. Experiments show that our method: is always more accurate than the other methods, and completes the analysis of the biggest available reference genome in about one day running at a rate of 0.98Gbp/h on a standard workstation.