Question

Use of pseudocounts to make a pseudo-sample

0

Entering edit mode

bli • 0

@bli-13334

Last seen 6.7 years ago

France

I see that estimateSizeFactors offers several options for determining the size factor. Some of them are meant to deal with the case where every gene contains a sample with a zero. I was wondering if the following would be a bad alternative to compute the pseudo-sample:

Add a pseudo-count of 1 everywhere.
Compute the geometric mean.
Substract 1.

Would this way of doing have bad properties?

DESeq2 • 1.1k views

ADD COMMENT • link updated 6.8 years ago by Michael Love 42k • written 6.8 years ago by bli • 0

score 0 · Answer 1 · 2017-07-31

0

Entering edit mode

Michael Love 42k

@mikelove

Last seen 4 hours ago

United States

You can try it out with simulated data, that's usually my approach. I can't think immediately of how this estimator would behave relative to e.g. "poscounts" for various types of data.

ADD COMMENT • link 6.8 years ago Michael Love 42k

0

Entering edit mode

Thanks for your quick reply.

About "poscounts", I read the following in the help:

          The "poscounts" estimator deals with a gene with
          some zeros, by calculating a modified geometric mean by
          taking the n-th root of the product of the non-zero counts.

Do I understand correctly that poscounts amounts to using a pseudo-count of 1 only for some genes, and only for the samples that have 0 for this gene (and not substracting 1 after computing the geometric mean)?