How to use lme4() to perform differential expression on bulk RNA seq data with multiple random effects?

Question

[bioyas]

Hi everyone,

I am dealing with bulk RNA-seq data that has one fixed effect (Condition) and two random effects (Patient info and batch info). I want to perform a DE analysis of the case and control conditions. Usually, I use the duplicateCorrelation built-in function to correct for random effects. However, this function can only handle one random effect. In my current analysis, I have more than one random effect.

I am aware of the Dream analysis from the variancePartition package. But for some reason, the analysis with this package fails. So I would like to proceed with lme4 package.

Is there any pipeline that goes over the differential expression analysis using lme4 package?

I would appreciate any help!

Thanks,

[ATpoint]

Is it not possible to make the batch a fixed effect? If you google "edgeR / limma and lme4" you will find some posts over at Bioconductor where the authors discourage to do that, mainly because the actual power of DE analysis (sharing information across genes to accurately estimate variance along the average expression gradient) is lost. Also, it's not standard so a reviewer might ask you to perform additional vaidation to confirm your DE analysis is proper.