Display Settings:

Items per page

Results: 19

1.
Figure 18

Figure 18. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of deletion lengths in our simulation.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
2.
Figure 19

Figure 19. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of deletion lengths detected with Sanger reads.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
3.
Figure 8

Figure 8. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Scatter plot of deletion lengths and differences of deletion calls. No correlation between deletion lengths and differences was observed (r2=0.056). ChopSticks worked well regardless of deletion lengths. Here, k=2, f=0.5, and coverage=5.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
4.
Figure 4

Figure 4. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Recall and precision of results of SV detection tools. BreakDancer and CLEVER achieved relatively good recall for all coverage, while recall of MoDIL was low. Although recall of CNVnator was not bad, its precision was low. The recall of an SR method Pindel was good when coverage was high, but it was insufficient when coverage was low.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
5.
Figure 7

Figure 7. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of differences of BreakDancer results and those improved by ChopSticks. The distribution of differences of ChopSticks results concentrated around zero, whereas that of BreakDancer results had long tail in 0–50 bp. Here, k=2, f=0.5, and coverage=5. Each frequency corresponds to the number of differences in bins of 2 bp.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
6.
Figure 9

Figure 9. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

MoDIL results improved by ChopSticks. Box-and-whisker plots of upstream differences of deletion calls obtained by MoDIL and those improved by ChopSticks. The format of this plot is exactly the same as that in Figure 6, except that results for coverage=15 were shown instead of those for coverage=20. The results in this figure indicate that ChopSticks can also improve the resolution of MoDIL results.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
7.
Figure 11

Figure 11. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of differences of CLEVER results and those improved by ChopSticks. The distribution of differences of CLEVER results had long tail in 0–50 bp, whereas that improved by ChopSticks concentrates around zero. Here, k=2, f=0.5, and coverage=5. Each frequency corresponds to the number of displacements in bins of 2 bp.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
8.
Figure 13

Figure 13. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Pindel results and those modified by ChopSticks. Box-and-whisker plots of upstream differences of deletion calls obtained by Pindel and those modified by ChopSticks. The format of this plot is exactly the same as in Figure 6. The results in this figure indicate that ChopSticks should not be applied to the Pindel results because the resolution of the Pindel results is already quite high.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
9.
Figure 10

Figure 10. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

CLEVER results improved by ChopSticks. Box-and-whisker plots of upstream differences of deletion calls obtained by CLEVER and those improved by ChopSticks. The differences were successfully corrected. Note that a significant portion of breakpoints predicted by CLEVER were inside the true deletion. Nonetheless, ChopSticks selectively trimmed predicted breakpoints outside true deletions, and left those inside untouched.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
10.
Figure 15

Figure 15. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of differences of BreakDancer results and those improved by ChopSticks. The distribution of differences of BreakDancer results had long tail in 0–400 bp, whereas that improved by ChopSticks concentrates around zero and frequencies in the long tail were reduced. Here, k=2, f=0.5. Each frequency corresponds to the number of differences in bins of 20 bp.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
11.
Figure 16

Figure 16. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Distribution of differences of CLEVER results and those improved by ChopSticks. ChopSticks corrected some of breakpoints predicted by CLEVER so that the peak at zero became stronger. However, the distribution of differences of CLEVER results had long tail in 0–3000 bp and it was difficult for ChopSticks to correct such large differences. Here, k=2, f=0.5. Each frequency corresponds to the number of differences in bins of 20 bp.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
12.
Figure 14

Figure 14. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

BreakDancer results for DBA/2J reads improved by ChopSticks. Box-and-whisker plots of upstream and downstream differences of deletion calls obtained by BreakDancer and those improved by ChopSticks. The results in this figure indicate that ChopSticks can improve the resolution of deletion calls for real sequences. Although ChopSticks trimmed upstream ends of a few deletion calls too much when k=1 or k=2 and f was small, such problems quickly disappeared for greater k and f values.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
13.
Figure 12

Figure 12. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

CNVnator results improved by ChopSticks. Box-and-whisker plots of upstream differences of deletion calls obtained by CNVnator and those improved by ChopSticks. The format of this plot is exactly the same as that in Figure 6. We expanded the original deletion calls of CNVnator outward by the window size (50 bp) because ChopSticks assumes that predicted breakpoints are outside true deletions. The results in this figure indicate that ChopSticks can improve the resolution of CNVnator results if predicted positions of breakpoints are within a few hundreds of bases from true breakpoints.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
14.
Figure 17

Figure 17. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Pseudocode of trimming algorithm. Pseudocode of the trimming algorithm of ChopSticks. Here, L is the length of the deletion call being processed, k is a threshold used to discriminate high-coverage regions from low-coverage ones, and f is a parameter that determines the threshold of the coverage of regions to be trimmed. The variable x represents the position of the base being examined, and the variable y represents the length of a region to be trimmed. The value c[x] is the coverage at the x-th base in the deletion call, while s keeps the sum of c[x] values.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
15.
Figure 3

Figure 3. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Overview of trimming algorithm of ChopSticks. Schematic illustration of the trimming algorithm of ChopSticks. ChopSticks trims ends of deletion calls that are not likely to be parts of deletions, according to their coverage. First, it trims high-coverage regions at the ends of deletion calls. Here, a high-coverage region is a region whose coverage is greater than a given parameter k. Second, it recognizes a high-coverage region separated by a low-coverage region and trims these regions if their joint coverage is deeper than kf, where f is another parameter. The second step is repeatedly conducted until the joint coverage becomes less than kf .

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
16.
Figure 2

Figure 2. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Expected resolutions of ChopSticks and threshold-based RP methods. The expected resolution of our method () is shown by a thick red line, that of threshold-based RP methods (E[Δb|b,c]) is shown by a thin solid black line, and that of threshold-based RP methods with double coverage (E[Δb|b,2c]) is shown by a dashed black line. The difference between and E[Δb|b,2c] is also shown by a dotted blue line. As the coverage goes away from zero, the resolution obtained by our method quickly outperforms that of normal RP methods. It is also clear that the resolution of our method is very close to that of threshold-based RP methods with double coverage. The difference approaches zero when coverage approaches zero or infinity, as indicated by the blue dotted line. E[Δb|b,c], , and E[Δb|b,2c] are given by Equations (2), (3), and (5), respectively. In this figure, d=200 and r=100.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
17.
Figure 5

Figure 5. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Number of deletion calls covering the whole of true deletions. Solid lines and circles show the number of all deletion calls generated by each tool, whereas dashed lines and ‘+’ symbol s show the number of deletion calls covering the whole of true deletions. Most of the deletion calls of MoDIL, CNVnator (expanded by the window size), and Pindel covered the whole of true deletions. On the other hand, many CLEVER results did not always contain the whole of true deletions, while median of the distribution of predicted breakpoints was close to the true breakpoints as shown in Figure 10. BreakDancer results for high coverage data did not always contain true deletions either. Predicted breakpoints of BreakDancer approached true breakpoints as the depth of coverage increases, and sometimes intruded into true deletions when coverage was high.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
18.
Figure 1

Figure 1. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

Resolution improvement by exploiting concordant read pairs. Schematic illustration of the key idea of our method ChopSticks. Unlike conventional SV detection methods based only on discordant pairs whose mapping distances were not close to the expectation, ChopSticks uses concordant read pairs as well. There is a chance that there is a concordant read closer to the boundary of the deleted region (breakpoint) than any discordant reads. Such a concordant read localizes the predicted position of the breakpoint, and therefore it contributes to achieving a high resolution. In this figure, b is the upstream end of a true deletion, Δb is the distance between the upstream end of a true deletion and that of a deletion call by threshold-based read-pair (RP) methods. Similarly, is defined for our method. The expected values of Δband are given by Equations (2) and (3), respectively.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.
19.
Figure 6

Figure 6. From: ChopSticks: High-resolution analysis of homozygous deletions by exploiting concordant read pairs.

BreakDancer results improved by ChopSticks. Box-and-whisker plots of upstream differences of deletion calls obtained by BreakDancer and those improved by ChopSticks. The red, green, blue, light blue, and magenta boxes correspond to k values of 1, 2, 3, 4, and 5, respectively, and the rightmost yellow box corresponds to the original results of BreakDancer. Among boxes of the same color, from left to right, f=0.1, 0.2, …, 1.0. Brown horizontal dashed lines indicate the values of 25%, 50%, and 75% tiles of differences of original deletion calls from below to above, respectively. The results in this figure indicate that ChopSticks clearly improved the resolution of the original BreakDancer results. When the coverage was low, small k values were effective in improving the resolution. When coverage was high, the resolution was also improved for large k values. Therefore, when the coverage is high, we recommend using large k values to avoid erroneous alignments of NGS reads and the genome. We omitted the results for coverage=15 because they were similar to those for coverage=20.

Tomohiro Yasuda, et al. BMC Bioinformatics. 2012;13:279-279.

Display Settings:

Items per page

Supplemental Content

Recent activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...
Write to the Help Desk