ChIP-seq data analysis for beginner

[ETHANol]

I should be getting back my first ChIP-seq data this week.

Which software do people recommend I use for data analysis, peak calling, etc. Any help with the criteria I should be asking to choose a certain software would be helpful and resources to answer my questions.

I'm bench scientist not a bioinforamtician so something reasonably user friendly would be nice, even better if I can run it on my MacBook.

Thanks.

[bioinfosm]

USEQ is a good tool with step-by-step guide

For windows solution, using cisGenome is good idea. It requires minimal pre-processing of solexa data before running it through cisGenome.

[ETHANol]

Thanks!!!! I'll try USEQ and report back in a week or so on how I fair.

[ETHANol]

USEQ looks like it has a lot of great tools, but is there more detailed information on how to use it. I'm really not understanding what I have to do to process my data.

[Chema76]

Hi,
If you are familar with R, perhaps one of these two packages will help you:
(try to use the last R version)

http://www.bioconductor.org/packages/2.6/bioc/html/CSAR.html

or

http://www.bioconductor.org/packages/2.6/bioc/html/PICS.html

We have developed a web page for the analysis of ChIP-seq data, now it is a beta version, but we will make it avaliable after this summer (only for plant genomes).

[ETHANol]

I checked out the Bioconductor apps and it seems you need to know R so I got absolutely nowhere with it.

Here's where I am. I have a sorted Illumina .txt files. I assume I need to convert this to a .BED file. Is there software that will do this for me in which I don't have to know a programing language.

Anybody have any suggested reading to get me up to speed. I'd really like to look at my data.

Thanks again.

[Chema76]

Once you have the file with the short reads, you should map them to the genome of interest, using SOAP, bowtie, Bwa or any other mapping tool.

The result of this mapping process should be used by the peak calling software (CSAR, PICS, Useq, Cisgenome...) to identify the significant regions.

With which organism are you working?
We have a webtool for the analysis of Arabidopsis ChIP-seq data. Basically, you submit the file with the short reads, and our server will analysis the data using SOAP and CSAR, it will report the binding map in a wig file, a the list of genes near by.

[czhang]

Use Starr from bioconductor. However you need know R basically.

[simonvh]

Have you looked at Galaxy? http://main.g2.bx.psu.edu/
I'm not really familiar with it as I prefer my trusty friend the command-line, but they have quite some nice tools, screencasts to explain typical analyses etc. It's all under active development and also specifically geared towards biologists.

[mceachin]

I have a related question. This is my first chip-seq analysis, 8 lanes of Solexa reads from a mouse experiment, and bowtie only aligns ~40% of the reads to mm9, no matter how stringent or lax I set the bowtie parameters. This alignment percentage seems low, compared to RNA-seq, but maybe it's not unusual for chip-seq.

Anyone with relevant experience, is this about right or should I be looking for an error?

Thanks

[simonvh]

I'm not familiar with Bowtie, but 40% seems quite low. The amount of reads should generally be higher than that. We routinely map 75%-85% of our ChIP-seq sample to mouse. Are reads mapping to repeat regions included in this number?

[Dethecor]

Quality Scores

I have seen things like that a couple of times when the quality scores of the reads were in a different scale than the default setting from bowtie, for example the bowtie manual states:

--phred33-quals
Input qualities are ASCII chars equal to the Phred quality plus 33. Default: on.

And my reads came from a newer solexa machine so i had to set --solexa1.3-quals which increased the percentage from ~40 to ~90% mapped reads in RNA-Seq Experiments.

This is because of the different scale a lot of good reads were discarded because their qualities were interpreted as being low when they actually were quite reasonable.

So you could try and check if your aligner discarded reads due to bad quality / if you used the correct quality scale.

Cheers

[mceachin]

Thanks, simonvh and Dethecor.

The unmapped reads are not obviously repetitive, but I'll check to see that the mm9 genome I'm aligning to is not masked for repetitive sequences. If that's the case, I'll try an unmasked genome.

In the mean time, I'm rerunning with the quality scale specified.

Thanks, mceachin

[Bioinfo]

Have you looked at Galaxy? http://main.g2.bx.psu.edu/
I'm not really familiar with it as I prefer my trusty friend the command-line, but they have quite some nice tools, screencasts to explain typical analyses etc. It's all under active development and also specifically geared towards biologists.

Hi Simon,
I am wandering that can we do two sample (Treated vs Control) in Galaxy.
thanks

[ETHANol]

Just a note on how I faired with my first ChIP-seq analysis for other beginners.

First I used the CLC genomics workbench as it's interface was really easy but ultimately was not satisfied with its performance.

I could never really get FindPeaks to work although I heard it's a great program.

After a little fiddling around I got USeq to work and so far I am very happy with it. The makers should be congratulated for producing a really nice package of programs. I really hope it is maintained. The ChIP-seq program wrapper doesn't work in my hands but that's no problem as it's probably better to process the data through the programs separately. The one thing that helped a lot getting it to work was when I found the "results > show results" menu option which tells you what went wrong when an error occurs.

The Galaxy MACS peak calling tool didn't recognize my Eland files so I never used it. But I did use my USeq peaks to map them to promoters as described in the webcast tutorial.