MACS fails, finds peaks but can't build model

Question: MACS fails, finds peaks but can't build model

3.1 years ago by

morgane.moreau • 20

Australia

morgane.moreau • 20 wrote:

Hi,

I'm using the Galaxy GVL-QLD instance.

I have single reads chip-seq files from E. coli. I set up my genome as a custom build.

I groomed, mapped my reads with BWA (when I use bowtie, some rows have the sequence in the quality score column and the quality scored in the OPT columns, and other rows are fine.. I got stuck there so I choose to map with BWA).

When I perform MACS (after removing unmapped reads via Filter SAM), down to MFOLD = 5 (which I know is not recommended but I'm testing the tool), it says that it finds 11 paired peak and it is not enough to built the model. And hence if fails.

I get this error message

INFO @ Tue, 27 Oct 2015 12:24:44: # ARGUMENTS LIST: # name = chip-seq_1_POS_MERGE # format = SAM # ChIP-seq file = /mnt/galaxy/files/000/117/dataset_117174.dat # control file = None # effective genome size = 3.22e+06 # tag size = 35 # band width = 75 #

I also get this error somtimes:

Traceback (most recent call last):
  File "/mnt/galaxy/shed_tools/toolshed.g2.bx.psu.edu/repos/devteam/macs/ae2ec275332a/macs/macs_wrapper.py", line 139, in <module>
    if __name__ == "__main__": main()
  File "/mnt/galaxy/shed_tools/toolshed.g2.bx.psu.edu/repos/devteam/macs/ae2ec275332a/macs/macs_wrapper.py", line 115, in main
    os.rmdir( treatment_dir )
OSError: [Errno 39] Directory not empty: '/mnt/tmp/953.1.all.q/tmplrIbXB/chip-seq_1_POS_MERGE_MACS_wiggle/treat'

I have tried to tweak the tag size (my reads are 51 bp) and the bandwidth, but it only increases the number of paired peaks (up to 21) but still fails. I am expecting only 6 peaks in my treatment sample.

Any suggestion to de-bug ?

Thank you :)

Morgane

galaxy instance macs fail • 904 views

ADD COMMENT • link •

modified 3.1 years ago by Jennifer Hillman Jackson ♦ 25k • written 3.1 years ago by morgane.moreau • 20

3.1 years ago by

Jennifer Hillman Jackson ♦ 25k

United States

Jennifer Hillman Jackson ♦ 25k wrote:

Hello,

A few suggestions:

1. Try using Bowtie instead of for mapping. With this analysis, where indels are located is not a factor (one of the differences between Bowtie and BWA output).

2. Tag size should be set to the same length as your reads.

3. Band-width appears to be on the narrow side. Try using the default (300) if you have not already, then tune it once a successful run is achieved.

It could be that the data is too shallow to call peaks, but these changes might help. Also know that you can filter peaks after the run, so if MACS calls more than the expected that is likely OK. Output all of the .wig files to help with that filtering plus visualization. An MFold of 6 is very low and likely not significant. More about MACS and parameters is online - a link to the full documentation is on the tool form.

Thanks, Jen, Galaxy team

ADD COMMENT • link written 3.1 years ago by Jennifer Hillman Jackson ♦ 25k

Thanks Jenny . Bowie messes up my Sam file as I mentioned in my first post . Any idea why is that so? Optimal Tag size and band width were used to start with but no peak were found . I am stuck. Would you be able to have a look at my workflow if u send you the Link ?

Thanks you

Morgane

ADD REPLY • link written 3.1 years ago by morgane.moreau • 20

Similar posts • Search »