Question: RNA-seq normalization methods for cross samples comparison?
gravatar for Hughie
3 months ago by
Hughie30 wrote:


Recently, I'm learning about RNA-seq quatification and I found there are many normalization methods, such as, CMP,RPKM/FPKM,TPM,TMM etc. After reading some papers about them, I noticed that all of these units mentioned above can't compare directly between samples because they just get relative expression instead of absolute expression.

I have just heard about spike-in method which may not be a proper method to get absolute expression, so, I want to ask you whether there exsits or you have imaged a proper expression unit to get absolute expression or compare between samples.

Thanks for your suggestion!

rna-seq sequence next-gen • 299 views
ADD COMMENTlink modified 3 months ago by kennethcondon2007840 • written 3 months ago by Hughie30

I believe spike-ins are used just to check that housekeeping genes are expressed in the correct proportions so are not appropriate for quantification.

TPM I believe is the most robust metric

As long as the samples come from the same experiment then they should be directly comparable because I would assume that the data has been generated the same way for each. Even if from different experiments, you would still use these metrics, but you would have to check that the experimental generation of the data was similar. You are far safer if you know that those that generated the data used spike-ins.

ADD REPLYlink written 3 months ago by kennethcondon2007840

Thank you! But, I'm thinking that TPM just measures relative expression, so ,if sampleA has a gene1 which has reads number much larger than the same gene1 of sampleB,(supoose 1000 vs 10),so,will this dramaticlly changed expression effect other genes because of 'relative'.

ADD REPLYlink written 3 months ago by Hughie30
gravatar for WouterDeCoster
3 months ago by
WouterDeCoster22k wrote:

compare between samples.

Do you want to do differential expression analysis? If so, the most appropriate tools are DESeq2 and edgeR. These will also deal with normalisation and provide the most sophisticated method so you don't have to worry about absolute vs relative quantification.

ADD COMMENTlink written 3 months ago by WouterDeCoster22k

Well, I see! Thank you !

ADD REPLYlink written 3 months ago by Hughie30

I have moved my reaction to an answer, so if this resolved your question you can mark it as accepted.

ADD REPLYlink written 3 months ago by WouterDeCoster22k

Sorry, I'm a beginner here, already accepted your kindly answer. Thank you again!

ADD REPLYlink written 3 months ago by Hughie30
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1308 users visited in the last hour