Questions about -cutoff and -desert options

[Magalie L]

Hello,

I have questions about the usage of desert and cutoff options. I've read the manual but did not find the information I need.

My meth files are from mapping and C count with the novoalign suite.

-desert option
What is the default value for this option in hmr and how to set it properly to deal with low coverage region between two good coverage regions?

How hmr makes the difference between no coverage and no cytosine ?

-cutoff in dmr
The defaut value for cutoff is 0.05, what does it mean?

How can I tune the c option to be more sensitive in regions with less difference between my samples, for example in comparison between same sample in different conditions or in other C contexts?


Thank you very much for your help,

Magalie

[Ben Decato]

Hi Magalie,

We have no difference between no coverage and no cytosine: these stretches of no information are deserts because we expect that there will be no correlation between CpG sites on either side of the desert, so it doesn't make sense to keep the Markov chain running.  Keeping the minimum desert size at the default (1000 bases) should work fine for varying coverage, but if your dataset is too fragmented (less than 40% of CpGs covered, for instance) the model may have difficulty tuning the parameters.

For -cutoff, that is ONLY used in dmr to count the number of "significant CpGs."  The p-value for each CpG is computed during the methdiff program using the one-directional Fisher's exact test.  With sufficient coverage, this p-value can be very low even for samples with very little methylation difference (as in different conditions of the same sample).  It is sometimes helpful to make a volcano plot for the CpGs with their p-value on the x-axis and methylation difference between the samples on the y-axis.

We usually recommend filtering DMR output with some kind of an awk script to get the "good quality" DMRs: those with some significantly different CpGs and by absolute methylation level difference with using the program roimethstat to calculate the DMR methylation level in each sample.

I hope this was helpful, feel free to mail back with any additional questions.  Thanks for using methpipe!!

Ben

--
You received this message because you are subscribed to the Google Groups "MethPipe and MethBase Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to methpipe+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

-- 
Benjamin Decato
PhD Student
Computational Biology Section,
Department of Biological Sciences,
University of Southern California
1050 Childs Way, RRI 408A
Los Angeles, CA 90089-2910

[Magalie L]

Dear Ben,

Thank you very much for all the details, it is very helpfull and I will follow your advice and try roimethstat on my samples.

Regards,

Magalie