Identical ChIP Seq pattern for different histone abs

[obischof]

We have performed ChIP-Seq using antibodies directed against different histone modfications. After bowtie and MACS14 treatment we find that all peaks fall into repeat regions. Worse, all abs give more or less the same peak pattern. Is this a problem of bowtie alignment parameters or MACS14??

[frozenlyse]

What were the modifications ie are the antibodies any good? Did you qPCR control regions (negative and positive) after the IP? How much yield did you get from the IP? How many PCR cycles did you do in the library prep PCR? How many reads did you get, and how many aligned? What parameters did you use for bowtie?

It sounds to me like you don't have any enrichment in your sequenced libraries - hopefully by asking these types of questions you can narrow down where something went wrong.

[jmw86069]

I agree with frozenlyse, and suggest running through some basic QC with something like the HOMER package. For example do you see clonal reads (symptom of PCR gone awry),GC bias (symptom of sequence composition issues), and do the reads map within a predictable distance (symptom of good/bad fragment length prep.) My first guess is that one of those measures will point to something "off."

[Simon Anders]

Have you used an input control? If not, your problem is sort of expected, because you will mainly see mapping artifacts and just open chromatin rather then true binding.

[frozenlyse]

Have you used an input control? If not, your problem is sort of expected, because you will mainly see mapping artifacts and just open chromatin rather then true binding.

It's only expected if the IP didn't work - Input controls are quite important but if you're not getting peaks in your IP, no amount of Input controls are going to change that... With the caveat that depending on which antibody/pulldown efficiency deciding what is an enriched peak/area may change.

[Simon Anders]

It's only expected if the IP didn't work - Input controls are quite important but if you're not getting peaks in your IP, no amount of Input controls are going to change that... With the caveat that depending on which antibody/pulldown efficiency deciding what is an enriched peak/area may change.

The OP wrote that all his peaks are the same across different samples, and then it would be interesting to see whether the input would have vetoed these.

I overlooked that he said they are all in non-mappable regions, i.e., he has no peaks at all, rather than, having many but all the same. Hence, you are right of course, most likely his IP did not work.

[mudshark]

did you run bowtie with -m 1 option to avoid repeat mappings?
which histone modifications did you analyze?

[obischof]

Our bowtie settings were as follows:

-a --best -q -m50 -e50 -solexa1.3-quals

We used a control input as reference and qPCRs performed on ChIP samples were all satisfactory for pos and neg controls. The library was amplified at 18 Cycles. The histone abs being used were all tested and passed that test.

[obischof]

Our bowtie settings were as follows:

-a --best -q -m50 -e50 -solexa1.3-quals

We used a control input as reference and qPCRs performed on ChIP samples were all satisfactory for pos and neg controls. The library was amplified at 18 Cycles. The histone abs being used were all tested and passed that test.

[mudshark]

would you mind running bowtie using only -m 1 and no other options?
how many reads did you get per sample?

[obischof]

I will do this asap. What does -m50 actually mean?? Would it not be better to run also with the -v parameter set eg at 2?

[mudshark]

I strongly suggest you check out the bowtie manual before specifying options.
http://bowtie-bio.sourceforge.net/manual.shtml

For ChIPSeq i would rather start with defaults and only limit the amount of reported hits to those which have just one reportable alignment (-m 1). this eliminates all hits to repetitive regions.