Question: TPM or FPKM for single cell RNA seq experiment ?
0
gravatar for aapapaiwannou
12 days ago by
aapapaiwannou0 wrote:

Hello,

I am doing my first NGS analysis and I am working with single cell RNA seq data. I want to be sure if its ok to use TPM for the downstream analysis (which I am planing to start with SINCERA pipeline )

thanks in advance for your help, Anna

rna-seq next-gen gene • 135 views
ADD COMMENTlink modified 12 days ago by Macspider3.0k • written 12 days ago by aapapaiwannou0
1

I would start here normalization in single cell RNAseq and from there keep digging forward. TPM/FPKM is a poor choice for inter-sample comparison. Please use the search function to understand why, we discussed this in numerous threads before.

ADD REPLYlink written 12 days ago by ATpoint24k

After some reading I should use TPM. So I have a count matrix( reads per transcript or fragments per transcript) derived from featureCounts (subread package) and I want to normalize in order to have TPM and afterwards to use this as an input to SINCERA pipeline for single cell transcriptome analysis

ADD REPLYlink written 12 days ago by aapapaiwannou0

I quote the same sentences that I quoted in the linked thread normalization in single cell RNAseq

Finally, library size normalization is another commonly used approach for normalizing RNA-seq data. This involves scaling the counts such that the library size is the same across libraries, and is the basis for measures of normalized expression like counts or transcripts per million. While simple, this approach is not robust to the presence of DE genes [3, 4]. This means that library size normalization is often inappropriate for real data sets in which DE is likely to occur.

Did you even read the linked thread and the paper that is suggested there? They show that TPM is a "suboptimal" choice.

ADD REPLYlink modified 12 days ago • written 12 days ago by ATpoint24k

neither of them is perfect.

ADD REPLYlink modified 12 days ago • written 12 days ago by shoujun.gu240
2
gravatar for Antonio R. Franco
12 days ago by
Spain. Universidad de Córdoba
Antonio R. Franco4.1k wrote:

It depends.. If you want to analyze differential expression with packages such as DESeq2 oe EdgeR, you need to provide the raw counts, since these packages has their own normalization methods

ADD COMMENTlink written 12 days ago by Antonio R. Franco4.1k
1

DESeq2 is not appropriate for single cell analysis.

ADD REPLYlink written 12 days ago by swbarnes26.7k

There are recommendations on how to use DESeq2 for single-cell applications which involves the zinbwave package. http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#recommendations-for-single-cell-analysis

ADD REPLYlink written 12 days ago by ATpoint24k
0
gravatar for Macspider
12 days ago by
Macspider3.0k
Vienna - BOKU
Macspider3.0k wrote:

The community is generally more used to FPKM but the author of FPKM itself said that we should move on because it is biased (ref).

Also, you can convert FPKM to TPM using a formula. And if you can't access that formula, I also report it in one of my papers (see methods section "ΔXT/FT expression profile").

I would really analyse both and take a conclusion after seeing both values for each gene.

ADD COMMENTlink written 12 days ago by Macspider3.0k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1421 users visited in the last hour