A ComBat-seq issue: the covariates are confounded
Entering edit mode
schlumbohm • 0
Last seen 5 weeks ago

Dear creators of ComBat-seq,

for experimental reasons, I was trying to run ComBat-seq on only two batches of the GFRN dataset provided (https://github.com/zhangyuqing/ComBat-seq/tree/master/real_data_application). I cropped the matrix accordingly (99 % sure it is not the issue). Everything works perfectly as long as I do not use the 'group' parameter. I made sure only to use the correct labels here. See my batch and group input:

batch <- factor(c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2))

group <- factor(c("gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "her2",
"her2", "her2", "her2", "her2", "gfp", "gfp", "gfp", "gfp", "gfp", "gfp", "egfr", "egfr",
"egfr", "egfr", "egfr", "egfr"))

Both have the same length of 29, which matches with my amount of samples/columns in the count dataset.

Be aware that I am a programmer myself and doublechecked everything to be correct. Everything should work, yet I get the following error:

The covariates are confounded! Please remove one or more of the covariates so the design is not confounded

Is there any hidden restriction on the covariates (groups) I use? With all three batches and respective metadata it works fine on the full GFRN dataset. In my example above, Batch 1 contains gfp and her2 and Batch 2 contains gfp and egfr.

I do need your input on this, I believe, since the chosen license prohibits me from altering and using it in a research and puclication setting.

Thank you in advance!

Simon Schlumbohm

ComBat-seq sva • 299 views
Entering edit mode

There is 29 values for batch and 47 for group, that cannot work. Please correct the question. Also, batch should be a factor, not an integer.

Entering edit mode

Dear ATpoint,

thank you for the comment. I did correct the question. For the group, I did accidentally share the input for all three batches, being 47 instead of 29. This was an error in the question posted here only, so it does not change anything regarding the outcome of my tests.

As for the batch, I can confirm, that a as.integer(c()) input works without a problem for the complete dataset, yet I have changed it in the question to factor as you suggested.


Login before adding your answer.

Traffic: 382 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6