Fig. 1

Method overview. a A sequence of scores: The score of each nucleotide is the adjusted count of the k-mer starting at this nucleotide. b Smoothed scores: The smoothed score is the weighted average of the flanking scores. The weights are assigned according to a Gaussian distribution. The local maxima, marked by ‘+’, are located using the second derivative test. c Candidate regions: The labeling module locates candidates (thin and colored in red) and potential non-repetitive regions (thick and colored in black). Regions found in the whole genome are used for training the hidden Markov model (HMM). d Final regions: The scanning module applies the trained HMM to locate the final repetitive regions (thin and colored in red). Notice that the final repetitive regions are less fragmented than the candidates. Additionally, they include all local maxima even the ones that were missed by the labeling module