How to decipher FPKM Values?

Question

[nikelle.petrillo]

Hello everyone,

I ran count_features_given_MIN_FPKM_threshold.pl to obtain some FPKM values on my transcript.

http://postimg.org/image/4tkh3pj13/

I have read through some articles in order to figure out what these FPKM values tell me, but i am only confusing myself more. Can anyone explain to me what these values mean / how they can be used to assess the quality of my transcript?

Thanks for your help! Nikelle

[Carlo Yague]

I looked at the script "count_features_given_MIN_FPKM_threshold.pl" and it seems that you misunderstood its purpose.The script counts the number of genes (features) that are truly expressed given an expression (FPKM) threshold. Running it "to obtain some FPKM values on my transcript" can't work. More info here.

Answer 1

[WouterDeCoster]

FPKM values is the amount of reads/Fragments per Kilobase of Transcript per Million reads. As such these numbers don't really tell you anything about quality.

[nikelle.petrillo]

Thank you, Do you know why FPKM is important for RNA seq?

[andrew.j.skelton73]

FPKM is a metric of quantification. It's mostly seen with the Tuxedo method of RNA seq analysis, TPM is widely considered a better metric to use. Ultimately I think that (normalised) counts are the best things to start with, as they're much easier to interpret.

[Carlo Yague]

FPKM is a metric used to measure gene expression. It is important in RNA-seq because you often want to identify difference of gene expression in such experiment.

[WouterDeCoster]

I would like to add that for DESeq2, a tool commonly used for differential expression analysis, raw counts are preferred over normalized counts.

[nikelle.petrillo]

Thanks very much, do you know how i would obtain these raw counts? I plan on using edgeR, and I believe that requires raw counts over normalized counts as well.

[nikelle.petrillo]

Thanks very much, do you know how i would obtain these raw counts? I plan on using edgeR, and I believe that requires raw counts over normalized counts as well.

[andrew.j.skelton73]

Use HTSeq_Count on each alignment (bam)

[Carlo Yague]

or featureCounts, which does more or less the same thing, although a bit faster.