Brandon:LabNotes/Project1/2012-2-1
Jump to navigation
Jump to search
Peakseq update[edit]
- generated data that output significant peaks. changed the thresholds in the config.dat file, target FDR of .10, max Q value of .10, and 50 simulations. It was not giving many significant peaks, 30 for 500 cells
- however when looking at data ranked by significance, any peaks past the 30th significant peak only had 2 reads in them
- running data without clonal read removal produced a 2000 ish significant peaks though....
UCSC and replicates comparison[edit]
- checked reproducibility between replicates in peaks called by peakseq by examining overlap between segments on base by base basis. Used coverageBed from bedtools which calculates overlap from bedfiles.
- only used got 60ish sites from UCSC data for comparisons.
$ coverageBed [OPTIONS] -a <BED/GFF/VCF> -b <BED/GFF/VCF>
- interpeak distance of 200; fdr q value .10
' | ' | ' | total bp | covered_bp | % covered |
1v2.txt | 1v2.txt | 500 cells v 200 cels | 6395 | 6718 | 95.19202143 |
1v3.txt | 1v3.txt | 500 cells v 100 cells | 6855 | 7047 | 97.27543636 |
2v3.txt | 2v3.txt | 200 cells v 100 cells | 4657 | 7047 | 66.08485881 |
1vUCSC.txt | 1vUCSC.txt | 500 cells v UCSC | 2115 | 9290 | 22.7664155 |
2vUCSC.txt | 2vUCSC.txt | 200 cells v UCSC | 1022 | 9290 | 11.00107643 |
3vUCSC.txt | 3vUCSC.txt | 100 cells v UCSC | 1263 | 9290 | 13.59526372 |
1v3-4.txt | 1v3-4.txt | 500 cells v 100 cells | 6855 | 7047 | 97.27543636 |
2v3-4.txt | 2v3-4.txt | 200 cells v 100 cells | 4657 | 7047 | 66.08485881 |
3-4vUCSC.txt | 3-4vUCSC.txt | 100 cells v UCSC | 1263 | 9290 | 13.59526372 |
- Comparing coverage from 500 cell sample to 100 and 200 cell samples, there is high reproducibility between replicates in terms of peaks called.
- not sure why comparison between 200 cells versus 100 cells does not have as much reproducibility as with 500 versus 100 and 200 cells. could be since its 500 cells, then its more likely to cover more regions that are also covered in 100 and 200 cells
- versus UCSC, lower concordance, which figures since it is a different cell line, and also a DNase I hypersensitivity method opposed to our tagmentation method.
- last 3 table entries used 500 cell control for 100 cell sample; for comparison against cell number matched controls. was no difference in the calling, thus the background normalization was the same in this case. normalizing the control to a different sample didn't result in different results, since background probably similar in them.
- tried increase interpeak distance to see if better calls would be made.
- interpeak distance of 1500; fdr q value .10
' | ' | total bp | covered_bp | % covered |
indx1vUCSC | 500 cells v UCSC | 2106 | 16369 | 12.86578288 |
Indx1v2 | 500 cells v 200 cels | 12226 | 33807 | 36.16410802 |
Indx2v1 | 200 cells v 500 cells | 12226 | 33807 | 36.16410802 |
Indx2v3 | 200 cells v 100 cells | 20361 | 24483 | 83.16382796 |
Indx3v1 | 100 cells v 500 cells | 10275 | 16369 | 62.7710917 |
- better calls not made. similarity between replicates went down. default and possibly lower interpeak distances = better