Difference between COHCAP.avg.by.site and COHCAP.avg.by.island

[Anonymous]

Hello,

I am doing some NGS methylation project with COHCAP and I am quite happy with it. Thank you.

I have a question though. I read the documents and ran COHCAP.avg.by.site and COHCAP.avg.by.island, but I am confused with which results I should use for my project.
what I want to know is differentially methylated islands, but it seems like both of the results have differentially methylated islands.

Here is my results.

Avg_by_Site
www.dropbox.com/s/zd5v9fmvw2gmn7o/CpG_Island_filtered-Avg_by_Site.JPG?dl=0

Avg_by_Island
https://www.dropbox.com/s/7wf4r0392haq8g2/CpG_Island_filtered-Avg_by_Island.JPG?dl=0

would you mind telling me the difference and which I should use for it?

Thank you,
Sorry for my English if it bothers you.
JK Kim

[Charles Warden]

Hi JK,

In most cases, I would recommend using the "Average-by-Island" results.

Both workflows have the same site-level analysis, but differ in how differentially methylated islands are defined from the differentially methylated sites.

The "Average-by-Island" workflow averages methylation values across differentially methylated sites within an annotated region (assuming that there are a minimum number of differentially methylated sites, which by default is 4). This defines one methylation value per sample, which is then tested among replicates for each group.

The "Average-by-Site" workflow averages methylation values per sites. The methylated and unmethylated thresholds (by default, 0.7 for methylated and 0.3 for unmethylated) is then used to assign a discrete methylation status based upon those average methylation values (and the ratio of methylated to unmethylated sites is compared within an annotated region). So, if you have a clear methylation difference, you can usually identify regions with either workflow. However, if you have heterogenous data (such as patient data), you may need to change the default thresholds (such as setting the methylated threshold to 0.3) and use the "Average-by-Island" workflow (if there are fewer average methylation values that meet the criteria to assign a methylation status to that site).

If you have gene expression data, you could also compare which strategy idenifies a larger percent of regions with negative associations with gene expression (with default parameters, this is probably the "Average-by-Island" workflow, but this would be one way to compare options with your own data).

Best Wishes,
Charles

[Anonymous]

Hello Charles,

Thank you so much for your answer.

Take care and Happy new year,
JK