Home

What is COHCAP?

COHCAP (Warden et al. 2013) is an algorithm to analyze single-nucleotide resolution methylation data. It provides QC metrics, differential methylation for CpG Sites, differential methylation for CpG Islands, integration with gene expression data, and visualization of methylation values.

How can I cite COHCAP?

Charles D. Warden, Heehyoung Lee, Joshua D. Tompkins, Xiaojin Li, Charles Wang, Arthur D. Riggs, Hua Yu, Richard Jove, Yate-Ching Yuan. (2013) COHCAP: An Integrative Genomic Pipeline for Single-Nucleotide Resolution DNA Methylation Analysis Nucleic Acids Research. 41 (11): e117

Tutorials

There are detailed tutorials in the COHCAP manual (available in the COHCAP package). You can also use the links below to download the relevant sections of the COHCAP manual:

Running COHCAP
Input Files
Output Files
Parameters
Visualizing .wig files using IGV or UCSC Genome Browser

Common Problems & Solutions

Problem: COHCAP cannot open a file
Possible Solution: Make sure that no files, folders in path to files, or groups have spaces. This will cause the program to crash.

Problem: COHCAP doesn't predict any differentially methylated CpG islands
Possible Solution: If you are working with patient data, it is likely to be more heterogeneous than cell line data. I would recommend setting the methyl.cutoff = 0.3 (instead of 0.7). This should increase the number of differenitally methylated sites and in turn increase the likelihood of seeing differentially methylated islands. If you are using the Bioconductor version, you'll want to set this parameter for both COHCAP.site() as well as COHCAP.avg.by.island() (or COHCAP.avg.by.site()).

Please click here to learn more about the methylated and unmethylated thresholds.

For the TCGA data, this was sufficient to produce useful results. If this is not sufficient for your own data, you can try the following strategies:

1) Increase the fdr.cutoff value to fdr.cutoff=0.25 (you can also set it to 1.00, but I'd recommend that as a last resort)

2) Focus on delta.beta and don't try to sites that say confined to the methylated and unmethylated peaks (by setting methyl.cutoff = 0 and unmethyl.cuffoff = 1)

3) You can change the num.sites value (default = 4), with 2 being the minimum number necessary to give you something different than the CpG site analysis.

4) Check your QC plots. Do the tumor and normal samples cluster together? Does it look like any of your samples are outliers? Your sample size is relatively small to begin with, but I would expect outlier removal to increase your ability to detect sites/islands.

Other FAQs

[Do you have a demo dataset?]
[What do I need to load into COHCAP for Illumina Methylation Array Data?]
[What do I need to do to prepare my BS-Seq data for analysis in COHCAP?]
[What is the format for the sample description file?]
[What is the format for the --param file?]
[What is the format for the gene expression file?]
[What are the command line parameters?]
[Why can't I run COHCAP GUI?]
[What if I don't use Genome Studio to produce my table of beta values?]
[Why doesn't my .wig file match the results in the COHCAP table for some UCSC CpG Islands?]
[I have pointed R to Rscript.exe in Windows - why doesn't COHCAP run properly?]
[What about CpG sites that are not located within UCSC CpG islands?]
[How do I know if Java JDK is installed correctly?]
[How do I create a new .jar file for my platform?]
[How do I know if Perl is installed correctly?]
[What if I do not have Illumina methylation array or targeted BS-Seq data?]
[What if I have 27k Illumina methylation array data?]
[What do the methylated and unmethylated thresholds mean?]

I still need help!

If the COHCAP user manual, FAQ links above, or discussion board cannot help you with your problem, you may contact Charles Warden (cwarden45 [at] gmail [dot] org]) for trouble-shooting support. However, we strongly encourage users to post their problems on the message board because they may be helpful to other users.