Long peak length from ChIP-seq data

Chiper · 12-25-2009, 10:30 PM

Hi All,
I am new to this forum. I have recently started using AB Solid V 2.0 for ChIP seq of a Drosophila TF. I used 500 bp avg Chromatin for pull down and for the library preparation it was brought down to ~100-200bp.

We ran MACS with default parameters on the data. I was surprised to see that mean peak length for Test IP vs.Input was ~1.6 Kb! (Similar results for Test IP Vs. Mock IP)

There were peaks as big as 10 Kb at times. Is this common? How can I get better resolution? Please help!

Xi Wang · 12-26-2009, 07:17 PM

what kind of ChIP-seq data did you use? if it is the histone modification data, the long peaks would not be surprise.

Chiper · 12-26-2009, 07:27 PM

Dear Wang,
Thanks for your reply. I didn't use histone modification ChIP-seq. The data is for a Drosophila transcription factor. Please let me know if you need more information.

Xi Wang · 12-27-2009, 06:04 PM

i didn't know much about the solid protocol for ChIP-seq. but i think the early stage software tools dealing with ChIP-seq data are for illumina data. check if the fragmentation of the solid protocol is the same as that of the illumina.

Chiper · 12-27-2009, 06:29 PM

Fragmentation of IP DNA (~500 bp) for Solid is done using Covaris S2 sonicator (water bath based, similar to Diagenode's bioruptor). I checked the size of the fragments on Lonza flesh gel after sonication and it was tightly distributed around 100-200 bp.

Xi Wang · 12-28-2009, 08:21 PM

The DNA fragmentation was carried out twice? I suggest that you could upload the raw reads and the MACS results to a genome browser to see how MACS works, and how MACS gives such long peaks.

Rao · 12-29-2009, 11:23 PM

I would suggest upload peaks and the reads that are responsible for these peaks to any genome browser. Compare the results and the distribution of reads in the peaks region. This may give a better idea.

Chiper · 01-08-2010, 04:47 AM

Thanks a lot for your comments. I could upload the mapped reads on UCSC browser. I could see good enrichment around known targets but there is noise in the data from adjacent regions. The data was subtracted with input and I am waiting for Mock subtracted data. Any suggestions on reducing such noise so that I can find smaller peaks? If I fail to get smaller peaks than is it OK if I use only those motifs which are near the peak summit?

Xi Wang · 01-08-2010, 05:00 AM

It is ok to look for motifs near the peak summit. but if there is no such consensus motif, you can extend the region to the flanking step by step, and do motif discovery again. To explain why no small peak, I am wondering if the TF of interest binds to the DNA sequence with its co-factors and the ChIP-ed DNA also contains the binding sites of the co-factors. Good luck!

Chipper · 01-08-2010, 07:05 AM

Chiper,

normally a size-selection is done prior to adaptor ligation and sequencing and one tries to keep fragments intact (i.e. no additional sequencing). The way your sample was treated means each fragment can give several reads, and reads from different strands will be less well separated. If your average size is 500 bp you may still have some enriched fragments that are several kb. One binding site could then give several "subpeaks" due to sonication or alignment bias. The resolution is not the same as the peak with though, the summits / peak max should still be what you use as a center for motif searches. You may find it difficult to separate close binding sites though.

Rao · 01-11-2010, 05:19 AM

Chiper,

If you see this paper at http://www.nature.com/nmeth/journal/...meth.1371.html, says three kinds of peaks, may give some idea about long peaks...

Classes of ChIP-seq signals. Consistent with previous ChIP-chip results, ChIP-seq tag enrichments or 'peaks' generated by typical experimental protocols can be classified into three major categories: punctate regions covering a few hundred base pairs or less; localized but broader regions of up to a few kilobases; and broad regions up to several hundred kilobases.

Also I will look for motif location in the binding sites...

Chiper · 01-12-2010, 12:43 AM

Thanks a lot Chipper and Rao for useful comments and the review link. I will look at the data again and write if there are other issues.

I love this community

Northeaster · 03-17-2010, 03:08 PM

Chiper,
Another solution is starting from ChIP: fragmentize the cross-linked sample to <500bp directly. Invitrogen has launched a SOLiD ChIP-seq kit, SOLiD™ ChIP-Seq Kit with ChIP Magnet, Cat. No. 4449638. In their manual, there are fragmentation condition using either bioruptor or covaris to shear chromatin into ~200-300bp fragments (actual range will be broader than this). For this case, you don't need 2nd covaris shearing for library construction. And, you will get better peak resolution.

12-25-2009, 10:30 PM	#1
Chiper Member Join Date: Dec 2009 Location: Asia Posts: 11	Long peak length from ChIP-seq data Hi All, I am new to this forum. I have recently started using AB Solid V 2.0 for ChIP seq of a Drosophila TF. I used 500 bp avg Chromatin for pull down and for the library preparation it was brought down to ~100-200bp. We ran MACS with default parameters on the data. I was surprised to see that mean peak length for Test IP vs.Input was ~1.6 Kb! (Similar results for Test IP Vs. Mock IP) There were peaks as big as 10 Kb at times. Is this common? How can I get better resolution? Please help!

12-26-2009, 07:17 PM	#2
Xi Wang Senior Member Join Date: Oct 2009 Location: Tsinghua, Beijing, China Posts: 236	what kind of ChIP-seq data did you use? if it is the histone modification data, the long peaks would not be surprise. __________________ Xi Wang

12-27-2009, 06:04 PM	#4
Xi Wang Senior Member Join Date: Oct 2009 Location: Tsinghua, Beijing, China Posts: 236	i didn't know much about the solid protocol for ChIP-seq. but i think the early stage software tools dealing with ChIP-seq data are for illumina data. check if the fragmentation of the solid protocol is the same as that of the illumina. __________________ Xi Wang

12-28-2009, 08:21 PM	#6
Xi Wang Senior Member Join Date: Oct 2009 Location: Tsinghua, Beijing, China Posts: 236	The DNA fragmentation was carried out twice? I suggest that you could upload the raw reads and the MACS results to a genome browser to see how MACS works, and how MACS gives such long peaks. __________________ Xi Wang

01-08-2010, 05:00 AM	#9
Xi Wang Senior Member Join Date: Oct 2009 Location: Tsinghua, Beijing, China Posts: 236	It is ok to look for motifs near the peak summit. but if there is no such consensus motif, you can extend the region to the flanking step by step, and do motif discovery again. To explain why no small peak, I am wondering if the TF of interest binds to the DNA sequence with its co-factors and the ChIP-ed DNA also contains the binding sites of the co-factors. Good luck! __________________ Xi Wang

12-26-2009, 07:27 PM	#3
Chiper Member Join Date: Dec 2009 Location: Asia Posts: 11	Dear Wang, Thanks for your reply. I didn't use histone modification ChIP-seq. The data is for a Drosophila transcription factor. Please let me know if you need more information.

12-27-2009, 06:29 PM	#5
Chiper Member Join Date: Dec 2009 Location: Asia Posts: 11	Fragmentation of IP DNA (~500 bp) for Solid is done using Covaris S2 sonicator (water bath based, similar to Diagenode's bioruptor). I checked the size of the fragments on Lonza flesh gel after sonication and it was tightly distributed around 100-200 bp.

12-29-2009, 11:23 PM	#7
Rao Member Join Date: Oct 2008 Location: USA Posts: 34	I would suggest upload peaks and the reads that are responsible for these peaks to any genome browser. Compare the results and the distribution of reads in the peaks region. This may give a better idea.

01-08-2010, 04:47 AM	#8
Chiper Member Join Date: Dec 2009 Location: Asia Posts: 11	Thanks a lot for your comments. I could upload the mapped reads on UCSC browser. I could see good enrichment around known targets but there is noise in the data from adjacent regions. The data was subtracted with input and I am waiting for Mock subtracted data. Any suggestions on reducing such noise so that I can find smaller peaks? If I fail to get smaller peaks than is it OK if I use only those motifs which are near the peak summit?

01-08-2010, 07:05 AM	#10
Chipper Senior Member Join Date: Mar 2008 Location: Uppsala, Sweden Posts: 161	Chiper, normally a size-selection is done prior to adaptor ligation and sequencing and one tries to keep fragments intact (i.e. no additional sequencing). The way your sample was treated means each fragment can give several reads, and reads from different strands will be less well separated. If your average size is 500 bp you may still have some enriched fragments that are several kb. One binding site could then give several "subpeaks" due to sonication or alignment bias. The resolution is not the same as the peak with though, the summits / peak max should still be what you use as a center for motif searches. You may find it difficult to separate close binding sites though.

01-11-2010, 05:19 AM	#11
Rao Member Join Date: Oct 2008 Location: USA Posts: 34	Chiper, If you see this paper at http://www.nature.com/nmeth/journal/...meth.1371.html, says three kinds of peaks, may give some idea about long peaks... Classes of ChIP-seq signals. Consistent with previous ChIP-chip results, ChIP-seq tag enrichments or 'peaks' generated by typical experimental protocols can be classified into three major categories: punctate regions covering a few hundred base pairs or less; localized but broader regions of up to a few kilobases; and broad regions up to several hundred kilobases. Also I will look for motif location in the binding sites...

01-12-2010, 12:43 AM	#12
Chiper Member Join Date: Dec 2009 Location: Asia Posts: 11	Thanks a lot Chipper and Rao for useful comments and the review link. I will look at the data again and write if there are other issues. I love this community

03-17-2010, 03:08 PM	#13
Northeaster Junior Member Join Date: Oct 2008 Location: california Posts: 2	Chiper, Another solution is starting from ChIP: fragmentize the cross-linked sample to <500bp directly. Invitrogen has launched a SOLiD ChIP-seq kit, SOLiD™ ChIP-Seq Kit with ChIP Magnet, Cat. No. 4449638. In their manual, there are fragmentation condition using either bioruptor or covaris to shear chromatin into ~200-300bp fragments (actual range will be broader than this). For this case, you don't need 2nd covaris shearing for library construction. And, you will get better peak resolution.