Different number of peaks when compared with ChIPseq ccbrpipeliner legacy version

kopardev commented 1 year ago

Description of the bug

For example: When using data from CCBR1187 (Legacy version output:/data/CCBR/projects/ccbr1187/ChIPseq_Pipeliner_220803), CHAMPAGNE version 0.2.0 is producing considerably lower number of peaks for all peak calling tools.

Command used and terminal output

No response

Relevant files

No response

System information

No response

kopardev commented 1 year ago

Sample: D1_STAT5_1: Legacy peaks: 15231; CHAMPAGNE peaks:126
Number of reads in the ingoing BAMs (surviving reads was very comparable between Legacy vs CHAMPAGNE).
loading MACS2 module and recalculating peaks with the same -t and -c BAMS yields: Legacy: 124 and CHAMPAGNE: 126. So, with newer version of MACS2, we are able to reproduce CHAMPAGNE peak calling results atleast for macs2 narrow peaks.

kopardev commented 1 year ago

creating new conda env with older version of MACS2 for testing .. the older version is no longer availabe on biowulf!
```
% conda create -n macs_2.1.1.20160309 python=2.7
% pip install pysam==0.9.0
% pip install numpy
% pip install MACS2==2.1.1.20160309
```
Testing using the same -t and -c BAMs as before with both Legacy (location:/data/CCBR/projects/ccbr1187/ChIPseq_Pipeliner_220803/tmp) and CHAMPAGNE (location:/data/CCBR/projects/ccbr1187/ChIPseq_Pipeliner_220803/tmp)

kelly-sovacool commented 1 year ago

This problem also happens when CHAMPAGNE is run on the ccbr1187 dataset with only the first read mates as if it were single-end.

sample_id	tool	count_old	count_new	rel_diff_percent
D1_P7+TLX3_STAT5_1	macs_broad	2223	258	-88.39
D1_P7+TLX3_STAT5_2	macs_broad	491	186	-62.12
D1_P7+TLX3_TLX3_1	macs_broad	113	11	-90.27
D1_P7+TLX3_TLX3_2	macs_broad	160	13	-91.88
D1_STAT5_1	macs_broad	18171	155	-99.15
D1_STAT5_2	macs_broad	392	153	-60.97
D1_TLX3_2	macs_broad	114	2	-98.25
D1_P7+TLX3_STAT5_1	macs_narrow	2075	231	-88.87
D1_P7+TLX3_STAT5_2	macs_narrow	467	172	-63.17
D1_P7+TLX3_TLX3_1	macs_narrow	114	7	-93.86
D1_P7+TLX3_TLX3_2	macs_narrow	145	9	-93.79
D1_STAT5_1	macs_narrow	15231	133	-99.13
D1_STAT5_2	macs_narrow	372	137	-63.17
D1_TLX3_2	macs_narrow	91	1	-98.90
D1_P7+TLX3_STAT5_1	sicer	43949	36604	-16.71
D1_P7+TLX3_STAT5_2	sicer	18063	12345	-31.66
D1_P7+TLX3_TLX3_1	sicer	17267	13082	-24.24
D1_P7+TLX3_TLX3_2	sicer	16871	16255	-3.65
D1_STAT5_1	sicer	45753	30657	-32.99
D1_STAT5_2	sicer	25164	13877	-44.85
D1_TLX3_1	sicer	31466	19671	-37.48
D1_TLX3_2	sicer	28301	14239	-49.69

(GEM is not shown here because it could not complete successfully due to an error message, no peaks were found).

See table for full paired-end data here: https://github.com/CCBR/CHAMPAGNE/issues/122#issuecomment-1788990782

kopardev commented 1 year ago

@kelly-sovacool I was able to generate peaks with my 1 sample run with GEM... thats interesting.

kopardev commented 1 year ago

@kelly-sovacool I was able to generate peaks with my 1 sample run with GEM... thats interesting.

Wait .. that was with PE input.

kelly-sovacool commented 1 year ago

After increasing macs q to 0.1, I'm still getting far fewer peaks than legacy pipeliner. Do we also need to tweak the broad_cutoff parameter? Workdir: /data/CCBR/projects/ccbr1187/champagne_2023-11-03

sample_id	tool	count_old	count_new	rel_diff_percent
D1_P7+TLX3_STAT5_1	gem	4337	2096	-51.67
D1_P7+TLX3_STAT5_2	gem	1303	690	-47.05
D1_P7+TLX3_TLX3_1	gem	1373	876	-36.20
D1_P7+TLX3_TLX3_2	gem	1281	632	-50.66
D1_STAT5_1	gem	2723	1324	-51.38
D1_STAT5_2	gem	1179	657	-44.27
D1_TLX3_1	gem	977	835	-14.53
D1_TLX3_2	gem	1190	782	-34.29
D1_P7+TLX3_STAT5_1	macs_broad	2223	210	-90.55
D1_P7+TLX3_STAT5_2	macs_broad	491	143	-70.88
D1_P7+TLX3_TLX3_1	macs_broad	113	22	-80.53
D1_P7+TLX3_TLX3_2	macs_broad	160	22	-86.25
D1_STAT5_1	macs_broad	18171	157	-99.14
D1_STAT5_2	macs_broad	392	131	-66.58
D1_TLX3_1	macs_broad	422	7	-98.34
D1_TLX3_2	macs_broad	114	7	-93.86
D1_P7+TLX3_STAT5_1	macs_narrow	2075	704	-66.07
D1_P7+TLX3_STAT5_2	macs_narrow	467	195	-58.24
D1_P7+TLX3_TLX3_1	macs_narrow	114	37	-67.54
D1_P7+TLX3_TLX3_2	macs_narrow	145	35	-75.86
D1_STAT5_1	macs_narrow	15231	1016	-93.33
D1_STAT5_2	macs_narrow	372	183	-50.81
D1_TLX3_1	macs_narrow	372	14	-96.24
D1_TLX3_2	macs_narrow	91	13	-85.71
D1_P7+TLX3_STAT5_1	sicer	43949	66942	52.32
D1_P7+TLX3_STAT5_2	sicer	18063	35309	95.48
D1_P7+TLX3_TLX3_1	sicer	17267	32213	86.56
D1_P7+TLX3_TLX3_2	sicer	16871	34259	103.06
D1_STAT5_1	sicer	45753	51428	12.40
D1_STAT5_2	sicer	25164	29108	15.67
D1_TLX3_1	sicer	31466	36199	15.04
D1_TLX3_2	sicer	28301	43853	54.95

CCBR / CHAMPAGNE