Question: methylPipe: how BSprepare handles zeros
gravatar for jonathan.moore
13 months ago by
jonathan.moore0 wrote:

I am looking at making mCG calls at single CpG sites, and have been looking at methylPipe for this reason.  I have a query about the functioning of BSprepare.

When BSprepare loads up the #C/#T values at a locus, if coverage<50 at a particular nucleotide, then it seems that it looks up the p-value of the site being methylated from a lookup table, which has been populated from binomial tests of #C+#T values from 1-50, and a supplied error probability.

The table has no entries for #C=0, and nor does BSprepare run a separate binomial test for #C=0 if #T<=50, instead returning NA.

Is it an intentional design feature that BSprepare returns NA as the p-value for cases where #C=0 and 0<=#T<50?  If so, is there a sound biological or statistical basis for this?

Many thanks for any help

Jay Moore

ADD COMMENTlink modified 13 months ago by mattia pelizzola200 • written 13 months ago by jonathan.moore0
gravatar for mattia pelizzola
13 months ago by
mattia pelizzola200 wrote:

Hi Jay,

we considered that in the absence of #C there is no evidence supporting the presence of an mC call at that position. Thus, we reasoned to be useless to test for it. You can consider this as a pre-processing filter, such as avoiding testing for differential expression for those genes that have either zero o very low expression.




ADD COMMENTlink written 13 months ago by mattia pelizzola200

Hi Mattia,

That makes great sense, thank you. It sounds like a useful performance optimisation.

Best wishes


ADD REPLYlink written 13 months ago by jonathan.moore0
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 247 users visited in the last hour