chip-exo data about macs HOT 11 CLOSED

Hi Cao,

Please try to run MACS2 using default parameter -- let it decide the fragment size. I tested some ChIP-exo data (human CTCF) before, and MACS can successfully find about 10bps fragment size for these datasets.

Give it a try, and let me know if it works.

Tao Liu

On Feb 19, 2013, at 10:57 PM, Cao Fan [email protected] wrote:

Hi,
I want to try MACS on chip-exo data. This kind of data differs from Chip-seq data in the sense that the peaks are sharper and more concentrated. And the distance between the w-peak and c-peak are smaller. Do you have any suggestions on how to tune the parameters for these data? Thanks.

Fan

—
Reply to this email directly or view it on GitHub.

from macs.

Hi Tao,
I tested both MACS 1.4 and MACS 2 on the chip-exo data with default settings, except for 1.4 I added --keep-dup auto. Then I looked at the models created by the two, it seems that they are very different. MACS1.4 is consistent across replicates, while MACS 2 gives some "weird" models. The attached are the models.
On the other hand, MACS2 appears to have a higher spatial resolution compared to MACS 1.4. So in this case, which version of MACS should be used? And are there any parameters to be tuned?
Thanks.

Fan

from macs.

Fan, did you use the version of MACS2 which can use cross-correlation to decide the fragment size? There must be another figure trying to find the lag between + and - tags to get the best correlation.

Tao Liu

Assistant Professor
Deptartment of Biochemistry
SUNY at Buffalo
NY State Center of Excellence in Bioinformatics & Life Sciences

B2-163 COEBLS
(O) 716-829-2749
[email protected]

Mailing address:
SUNY at Buffalo-COEBLS
701 Ellicott St, B2-163
Buffalo, NY 14203-1221

On Feb 20, 2013, at 2:49 AM, Cao Fan [email protected] wrote:

Hi Tao,
I tested both MACS 1.4 and MACS 2 on the chip-exo data with default settings, except for 1.4 I added --keep-dup auto. Then I looked at the models created by the two, it seems that they are very different. MACS1.4 is consistent across replicates, while MACS 2 gives some "weird" models. The attached are the models.
On the other hand, MACS2 appears to have a higher spatial resolution compared to MACS 1.4. So in this case, which version of MACS should be used? And are there any parameters to be tuned?
Thanks.

Fan

—
Reply to this email directly or view it on GitHub.

from macs.

Yes, I used the version with cross-correlation. The attached are the cross-correlation plots. The fragment size differ by a lot in the two replicates.

from macs.

To me, 124 seems like from the size selection, which means it reflects more on noise than real signal. The second one, 18bps looks typical for a ChIP-exo dataset.

Tao

On Feb 20, 2013, at 10:25 PM, Cao Fan [email protected] wrote:

Yes, I used the version with cross-correlation. The attached are the cross-correlation plots. The fragment size differ by a lot in the two replicates.

—
Reply to this email directly or view it on GitHub.

Tao Liu

Assistant Professor
Deptartment of Biochemistry
SUNY at Buffalo
NY State Center of Excellence in Bioinformatics & Life Sciences

B2-163 COEBLS
(O) 716-829-2749
[email protected]

Mailing address:
SUNY at Buffalo-COEBLS
701 Ellicott St, B2-163
Buffalo, NY 14203-1221

from macs.

Thanks for the comment. For MACS, is it possible to output the forward and reverse strand paired peaks that make up the peak called?

from macs.

Perhaps I can add it to 'predictd' module...

from macs.

Hi Tao,

I found this thread and have some questions about ChIP-exo.

For ChIP-seq data, as far as I understand, MACS1 extends the reads to 200bp, and then creates a wig file. However, MACS2 no longer output a wig file.

How can I get a wig file(or bedgraph) from ChIP-exo for visualization? I can generate a bedgraph for minus and plus strand respectively by bedtools(genomecov) or Homer, but the real peak will be in the middle of the paired minus and plus peaks.

Thanks so much!

Ming Tang

from macs.

bedGraph from MACS2 is better than wig from MACS 1 since it has 1-basepair resolution. In order to visualize it, convert bedGraph to bigwig and load it to IGV or UCSC genome browser. Bedgraph in MACS2 is generated by the same way as in MACS1, extend tag towards 3' direction with a fixed extension size then pileup. So the highest point of a peak region is the center of binding site.

Tao

On Nov 6, 2013, at 11:43 PM, Ming Tang [email protected] wrote:

Hi Tao,

I found this thread and have some questions about ChIP-exo.

For ChIP-seq data, as far as I understand, MACS1 extends the reads to 200bp, and then creates a wig file. However, MACS2 no longer output a wig file.

How can I get a wig file(or bedgraph) from ChIP-exo for visualization? I can generate a bedgraph for minus and plus strand respectively by bedtools(genomecov) or Homer, but the real peak will be in the middle of the paired minus and plus peaks.

Thanks so much!

Ming Tang

—
Reply to this email directly or view it on GitHub.

Tao Liu

Assistant Professor
Department of Biochemistry
University at Buffalo
NY State Center of Excellence in Bioinformatics & Life Sciences

B2-163 COEBLS
(O) 716-829-2749
[email protected]
http://biomisc.org/

Mailing address:
University at Buffalo-COEBLS
701 Ellicott St, B2-163
Buffalo, NY 14203-1221

from macs.

Thank you so much for your reply! I am running MACS1 for my ChIP-exo ( the computing cluster only have MACS1 installed, I can request MACS2 to be installed but it takes time). MACS1 and MACS2 should be both suitable for ChIP-exo right?

one other question is that I have a IgG control rather than an input control, does it matter?
and does MACS1 normalize the counts to counts per million mapped (CPM)?
I want to compare my TF binding in two different conditions. If I load the resulted bedgraph file to IGV,
can I directly compare the peak height?

Also, after running MACS two bedgraph files are given, one for IgG control, one for IP sample. can MACS generate a bedgraph file that subtracted the IgG background?

Thank you again!

Ming Tang

from macs.

Closing due to lack of activity.

from macs.