FindPeaks 3.1.3 (alpha) - ChIP-Seq and UCSC-compatible short-read wig track creator

[apfejes]

Hi All,

Just thought I'd take a second to let people know that FindPeaks 3.1.3 has been released. It's currently alpha code, so there are probably still a few bugs remaining, but since I've been running the code myself, it seems to do what I expect it to do.

In any case, if you'd like to see some of the improvements over FindPeaks 2.1.3 (the last version available on the web), I've put an abbreviated version of the change log on the download page.

http://www.bcgsc.ca/platform/bioinfo/software/findpeaks/releases/3.1.3

Happy ChIP-Seq-ing!

Anthony Fejes

[bioinfosm]

any comparisons with the Solexa's BeadStudio chipSEQ module?

I am looking to search for peaks on the whole human genome reference for a TF

[apfejes]

I have not tried out the Illumina Chip-Seq module - I don't use the Windows platform, so I probably won't get around to trying it out, either.

We tend to move very quickly to develop new features and fix bugs with our in-house version, which makes it an excellent tool for trying to squeeze out new information from ChIP-Seq experiments. It's also developed to be useful for both pipeline processing and for desktop use, whereas the Illumina version appears to be very GUI-centric. They're likely developed for different audiences.

As for your goal of trying to search for peaks on the whole human genome reference with a transcription factor - All chip-Seq programs can do that. FindPeaks, the Wold Lab's version (and others as well) and I'd be very surprised if the Illumina tool couldn't do it.

The underlying issue is what methods are available for finding Peaks in the package you select. Not all peaks contain the same amount of information, and how you extract and process that information is important.

[bioinfosm]

I see that findpeaks depends on the unique hits from Eland. But one can run eland with any seed length (default 32). Have you done tests to see what could be a better option. Using a lower threshold for eland like 25 to get more input data to find peaks...?

thanks !

[apfejes]

Hey Bioinfosm,

You're right, you can get better/worse results depending on the length from Eland. What we normally do is to run it at several lengths (e.g, 32, 29, 26, 23) and then use the ParseMultipleElands.jar application to pick the best hit from the longest Eland run. This way, you get the most specific hit, without losing too much information. The best of all possible worlds.

I once had the opportunity to take a look at 51-mers from a (Non-chip-seq) solexa run (last June, so the quality wasn't great), and found that the vast majority of hits came from 32-mers or greater, but there were still a significant number of 21- and 23-mers that were recovered by doing the shorter align lengths, about 3M and 3.25M respectively, out of a total of 30M reads. (Interestingly enough, there were about 3 Million 51-mers that aligned for the full sequence length, but thats another subject.)

Cheers,

Anthony