Question

how to remove batch effects of RNA seq data based on continuous variables(like surrogate variables from sva package) using combatseq or limma (removebatcheffect function)?)

1

Entering edit mode

amnahsiddiqa ▴ 10

@amnahsiddiqa-15854

Last seen 16 months ago

United States

Hi All,

I have counts normalized matrix and sva run on it and have two surrogate variables (which are continuous variables ofcourse) for hidden batch effects in my RNAseq data.

Now I am aware of combatseq package and limmas removebatcheffect() function but they don't take continuous variables. What are my options or probably I am missing something regarding continuous batch variables(in this case surrogate variables from sva). Please help me.

combatseq RNASeqData sva limma • 1.6k views

ADD COMMENT • link 17 months ago amnahsiddiqa ▴ 10

score 1 · Answer 1 · 2023-06-30

1

Entering edit mode

Gordon Smyth 51k

@gordon-smyth

Last seen 24 minutes ago

WEHI, Melbourne, Australia

removeBatchEffect does take continuous variables. It works with any combination of discrete factors and continuous covariates.

If your purpose is a differential expression analysis, then you don't even need batch removal, you simply include the surrogate variables in the design matrix.

ADD COMMENT • link 17 months ago Gordon Smyth 51k

1

Entering edit mode

@OP, it's the covariates argument in said function.

ADD REPLY • link 17 months ago ATpoint ★ 4.5k

0

Entering edit mode

Thank you both- Yeah I figured covariates is my shot to go. Just for everyone else who comes back , code below worked for me without any problem and my batch effect was certainly controlled for that I was able to visualize with pac :)

normalized_data <- counts(dds, normalized = TRUE)%>%data.frame(check.names = FALSE)
normalized_data_log=log2(normalized_data+1)
design <- model.matrix(~1+visit+response_status,metadata)
normalized_data_log_batcheffectsremoved <- removeBatchEffect(normalized_data_log,covariates =metadata$sv1, design=design)

ADD REPLY • link 17 months ago amnahsiddiqa ▴ 10