I have several RNA-seq raw counts datasets that I want to normalize to TPM. I exctracted the transcript lengths from ensemble, this is for example the lengths for one of the datasets, and as you can see the gene names are the rownames:
Now according to this awesome article, it is better to use effective lengths rather than just the transcript lengths, which is the gene length - X + 1, where X is "mean of the fragment length distribution which was learned from the aligned read".
My question is how do I calculate that in R?
What exactly do they mean with this X, and is the gene length the same as the transcript length? I'm a bit confused.
1 answer
You don't, at least not starting from a count matrix. If you read the linked blog carefully you would see that calculation of effective gene length requires information such as about fragment length distribution which is not available from such a matrix. You need a dedicated tool to calculate that starting from the reads itself, popular options are salmon, kallisto, or RSEM. That having said, you would need to start from scratch with the fastq files. With only counts you cannot achieve that.
Log in to answer this question.
-
Unexpected negative correlation between gene-length and counts
written by Ben •
-
Recommended way to normalize SmartSeq2 gene expression matrix to better match 10X expression data
written by Cookin •
-
Understanding log2 transformation for TPM
written by AlexStar •
-
The best way to get gene lengths for 15K+ genes?
written by AlexStar •
-
Warning - longer object length is not a multiple of shorter object length
written by AlexStar •
-
How to get gene lengths ?
written by AlexStar •
-
TPM for scRNAseq
written by little_more •
-
Calculating FPKM and TPM by hand from htseq-count output?
written by Joel Wallenius
-
Effective length in normalisation for reads counts
written by Xiaokang •
-
How do you normalize Transcript per Million TPM to compare between samples ?
written by ZheFrench
