Using RRBS and oxo-RRBS data with Merge Methcounts and Radmeth Adjust

[Chris S]

Hi, 

My data on mouse DNA was collected using RRBS methods to reduce the sequence depth needed.

Using merge methcounts, I combined the 8 samples as described in the manual. However, since I have RRBS a large portion of the file is zeroes for both reads and coverage. It seems reasonable to me to remove rows where all samples had zero for coverage and methylated reads using awk. 

Similarly, the Radmeth adjust feature is using multiple adjacent CpGs, which doesn't happen very often in RRBS data. Would it be more appropriate to filter out the rows of zeroes due to experimental design or to leave them in for the statistical calculations. My thought for removing them is to reduce the number of lines to keep the FDR threshold lower. 

Thanks,

Chris

[Ben Decato]

Hi Chris,

Thanks for using our pipeline. Radmeth requires the same set of CpGs present in the proportion table for all samples, but there is no rule on which CpG sites those are or whether they are even single CpG sites. For example, I have computed average methylation level in gene promoters and then replaced the CpG sites in the proportion table with the ~25k promoter methylation levels instead to reduce the number of statistical tests.

If you go this route (as I probably would with RRBS data), it probably isn't a good idea to use radmeth's adjust/merge functions. Run radmeth regression and then adjust the p-values using something like the p.adjust function in R, since the "adjacent" sites could be very far apart and therefore are likely to be independent.

I hope this helps!

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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

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

[Chris S]

Thanks Ben!

On Monday, June 25, 2018 at 1:34:34 PM UTC-5, Ben Decato wrote:

Hi Chris,

Thanks for using our pipeline. Radmeth requires the same set of CpGs present in the proportion table for all samples, but there is no rule on which CpG sites those are or whether they are even single CpG sites. For example, I have computed average methylation level in gene promoters and then replaced the CpG sites in the proportion table with the ~25k promoter methylation levels instead to reduce the number of statistical tests.

If you go this route (as I probably would with RRBS data), it probably isn't a good idea to use radmeth's adjust/merge functions. Run radmeth regression and then adjust the p-values using something like the p.adjust function in R, since the "adjacent" sites could be very far apart and therefore are likely to be independent.

I hope this helps!

Ben

On Mon, Jun 25, 2018 at 8:27 AM, Chris S <clseil...@gmail.com> wrote:

Hi, 

My data on mouse DNA was collected using RRBS methods to reduce the sequence depth needed.

Using merge methcounts, I combined the 8 samples as described in the manual. However, since I have RRBS a large portion of the file is zeroes for both reads and coverage. It seems reasonable to me to remove rows where all samples had zero for coverage and methylated reads using awk. 

Similarly, the Radmeth adjust feature is using multiple adjacent CpGs, which doesn't happen very often in RRBS data. Would it be more appropriate to filter out the rows of zeroes due to experimental design or to leave them in for the statistical calculations. My thought for removing them is to reduce the number of lines to keep the FDR threshold lower. 

Thanks,

Chris

--
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.