Different number of peaks when compared with ChIPseq ccbrpipeliner legacy version

[kopardev]

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]

• 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]

• 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]

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: #122 (comment)

[kopardev]

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

[kopardev]

@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]

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