help with bdgdiff

[Venu Pullabhatla]

Hi Tao

I have now managed to run bdgdiff from MACS_2.0.10 after following your
suggestions and now trying to interpret the output. I get 3 files as
output which look like (diff_ER_KO1_c1.0_cat1_peaks.encodePeak,
diff_ER_KO1_c1.0_cat2_peaks.encodePeak and
diff_ER_KO1_c1.0_cat3_peaks.encodePeak), but not the files which I
expect like consistent and unique sites. I got the unique and consistent
sites outputs while using macs2diff from version 2.0.9.

Before running "macs2 bdgdiff" I ran "macs2 callpeak" with --SPMR option
and got the bedgraph tracks.

here is how I ran bdgdiff for differential binding using the bedgraph
tracks from "macs2 callpeak"

/home/venu/tools/MACS_2.0.10/bin/macs2 bdgdiff --t1 ER_treat_pileup.bdg
--t2 KO1_treat_pileup.bdg --c1 ER_control_lambda.bdg --c2
KO1_control_lambda.bdg -o diff_ER_KO1

The files I used as inputs are the bedgraph files (_treat_pileup.bdg and
_control_lambda.bdg) from "macs2 callpeak" between ER + control and KO1
+ control as suggested by "macs2 bdgdiff -h"


-h, --help show this help message and exit
--t1 T1BDG MACS pileup bedGraph for condition 1. REQUIRED
--t2 T2BDG MACS pileup bedGraph for condition 2. REQUIRED
--c1 C1BDG MACS control lambda bedGraph for condition 1.
REQUIRED
--c2 C2BDG MACS control lambda bedGraph for condition 2.
REQUIRED
-C CUTOFF, --cutoff CUTOFF
Cutoff of log10 likelihood ratio cutoff.
DEFAULT: 1.0
(likelihood ratio=10)
-m MINLEN, --min-len MINLEN
Minimum length of differential region. DEFAULT: 200
-o OPREFIX, --o-prefix OPREFIX
output file prefix, DEFAULT: diffpeak
--depth1 DEPTH1 Sequence depth of condition 1 in million reads.
Default: 1 (if you use 'macs2 callpeak --SPMR' to
generate bdg files.
--depth2 DEPTH2 Sequence depth of condition 2 in million reads.
Default: 1 (if you use 'macs2 callpeak --SPMR' to
generate bdg files.

Is it right to use these bedgraph files or shoud I be using the bedgraph
tracks from "bdgcmp" (using p/q-value or Fold change etc) between T1+C1,
T2+C2 and T1+T2, T2+T1.

Many Many thanks for your help

Best wishes

Venu

[Venu Pullabhatla]

Hi

I did not succeed with bdgdiff, but was quite happy with the results we got from macs2diff from 2.0.9. It clearly produced the regions which are unique in both conditions and consistent in both. It was only possible because we have access a highmem machine which could hold 120m reads per condition. As Tao suggested macs2diff needs lot of memory (~30GB for two conditions with 30m non-redundant reads each)

I used the coordinates from the unique sites to do the pileup and could also calculate the fold change.

I am still looking to find solution for the problem with bdgdiff in 2.0.10.

Thank you



On 13/12/12 11:12, TRypdal wrote:

Hi,

I'm also interested in knowing about this, could anybody help? Venu, have you managed to find to proper way of doing this and if so, would you be able to help?

Thank you!

-- 
Venu Pullabhatla, Ph.D.
Bioinformatician
NIHR GSTFT/KCL Comprehensive Biomedical Research Centre
Guy's & St. Thomas' NHS Foundation Trust
8th Floor, Tower Wing, Guy's Hospital
Great Maze Pond, London SE1 9RT