Fig 1.
Illustration of the k-mer counting.
Long sequence: a sequence read. Length-3 sequence: k-mer. Here, k = 3. The table on the right shows the k-mer counting result.
Fig 2.
Illustration of the main steps of REPdenovo.
Thick bars: genomic sequences. Thin bars: k-mers. K-mer counting step: yellow parts are repeats (with some mismatches). Colored squares within thick bars: mutations (substitutions and indels) within repeats.
Table 1.
Sequence reads information from four human individuals from the 1000 Genomes Project.
# of reads: in millions. Coverage: average sequence depth per base.
Table 2.
The number of classified repeats constructed by REPdenovo on four different human individuals for fK = 10 and 100.
Classified into: (i) mappable to the reference genome, (ii) unmappable to the reference but have NCBI Blastn hits, and (iii) unmappable to the reference and no NCBI Blastn hits. The repeats in (ii) and (iii) may potentially be previously unknown repeats.
Fig 3.
Distribution of repeat matching lengths relative to their total length for fK = 10 and fK = 100.
Solid bars: repeats mappable to the reference genome. Bars with patterns: repeats unmapped to reference and having NCBI Blastn hits. The figure shows the relative matching length as the mapping ratio (0%-100%), which is the ratio between the length of mapped part and total length of the repeat. A majority of constructed repeats can match fully to the reference genome or have NCBI Blastn hits.
Fig 4.
Hits of Repbase repeats found by REPdenovo.
X axis: divergence rate (mismatches per 1,000 bases) of repeats given by Repbase. Y axis: number of copies from the UCSC genome browser annotation. Dots: Repbase repeats. Red dots: hits found by REPdenovo. Blue dots: repeats not found by REPdenovo.
Table 3.
Numbers of repeats that hit Repbase (with matching cutoff 0.0) and masked by RepeatMasker (with matching cutoff 0.0).
We classify the repeats based on whether they are mappable to the human reference and whether they have matches in Repbase. Masked: RepeatMasker can classify the repeat. Unmasked: RepeatMasker cannot classify the repeat.
Table 4.
Classification of the 190 un-mappable repeats with NCBI Blastn hits.
The numbers in parentheses are the numbers of repeats in each category.
Table 5.
Length distribution of the 190 potentially novel repeats.
The numbers in parentheses are the numbers of repeats in each category. For each range of repeat lengths, the number is the percentage of repeats falling in the range.
Table 6.
Classification of the 26 (out of the 190 potentially novel repeats) repeats that have no Blast hits on long reads reference.
The numbers in parentheses are the numbers of repeats in each category.
Table 7.
Assembly quality comparison of REPdenovo and RepARK.
N: the number of assembled contigs. Nh: the number of complete Repbase hits from the N repeats (with 85% coverage cutoff). : average coverage of hits. Cm: maximum coverage of hits by single assembled repeats. N50: N50 of assembled repeats.
Fig 5.
Assembled repeats matching AluYd3 (a Repbase repeat) by REPdenovo (bottom panel) and RepARK (top panel).
The matched assembled repeats are shown on their mapped positions where the AluYd3 consensus repeat sequence serves as the reference.
Table 8.
The number of repeats in Repbase that match (over the minimum threshold tL) one de novo repeat.
The numbers outside and side the parentheses are REPdenovo and RepARK results, respectively. tL: matching cutoff.
Table 9.
The number of repeats in Repbase that match (over the minimum threshold 0.85) one de novo repeat for different k-mer length LK.
By default, REPdenovo use different k-mer length (29, 39, and 49) together, and its result is marked as “Combined”.