Question

technical replicates and normalization/log transformation

0

Entering edit mode

A ▴ 40

@a-14337

Last seen 4 months ago

United Kingdom

hi,

I have a question regarding technical replicates of the same biological sample.

I have for a sample the following:

sample1-rep1
sample1-rep2

Sample2-rep1
Sample2-rep2
etc up to 12

The individual samples are biological replicates while the the sample (rep1 and 2) are technical replicates from the same tissue.

My question is, for DESeq2, should this cause issues for normalization/VST transformation? I primarily need the normalize counts/vst transformed counts for further downstream analysis (not DE)

for my model effect, batch is treated according to when library preps took place.

please note, i do not intend to collapse technical replicates.

thanks!

deseq2 normalization • 979 views

ADD COMMENT • link updated 3.6 years ago by Michael Love 41k • written 3.6 years ago by A ▴ 40

score 0 · Answer 1 · 2020-09-18

0

Entering edit mode

Michael Love 41k

@mikelove

Last seen 1 day ago

United States

We do recommend to collapse the technical replicates if you want to examine or test on the biological differences.

I don't see a problem having the technical replicates in for a VST PCA plot though.

ADD COMMENT • link 3.6 years ago Michael Love 41k

0

Entering edit mode

great! thank you for this advice.

As a follow up, is there a way to generate normalized counts without running the model? For example, would this work?

deObject<- DESeqfrom...etc
deObject<-estimateSizeFactors(deObject)
df<-counts(deObject, normalized = T)

and

vsd <- vst(dds, blind = T)

mat <- assay(vsd)
mat <- limma::removeBatchEffect(mat, vsd$batch)
assay(vsd) <- mat
write.csv(assay(vsd etc..)

Presumably the model does not need to be run in order to generate normalized counts as above? thank you again

ADD REPLY • link 3.6 years ago A ▴ 40

1

Entering edit mode

Yes, normalized=TRUE only requires estimateSizeFactors to be run.

ADD REPLY • link 3.6 years ago Michael Love 41k

0

Entering edit mode

thank you very much for your time :)

ADD REPLY • link 3.6 years ago A ▴ 40