FPKM threshold and filter

Question

[qudrat]

Can somebody suggest me how to choose the minimum FPKM value as a threshold to filter low quality transcripts?

[popayekid55]

you may choose 1 fpkm as cutoff for a trancript/gene to be expressed. anything below that could be filtered out

[Renesh]

Filter out all transcripts with FPKM variance less than one across all samples/conditions

[WouterDeCoster]

That highly depends on what you want to do with the data afterwards.

[qudrat]

My real aim is to find novel genes and many genes show very less value of FPKM.

[WouterDeCoster]

I obviously don't know how well your organism of interest has been characterized, but depending on what you are studying (organism, tissue) these 'novel genes' will probably have a low expression. So throwing away lowly expressed genes is not what you want to do. Another criterion would be (if it's eukaryotic) to see if it's a spliced transcript, or if it corresponds to a conserved sequence in the genome.

You'll probably need a lot of sequencing data, and perform validation afterwards to confirm you are not looking at an artefact.

Answer 1

[igor]

You can try to use a zFPKM transform to define expressed genes: https://bioconductor.org/packages/devel/bioc/vignettes/zFPKM/inst/doc/zFPKM.html

we provide a novel normalization metric, zFPKM, that identifies the threshold between active and background gene expression; and we show that this threshold is robust to experimental and analytical variations

Answer 2

[BIOTECH.DEEPTI911]

Hi folks, I am having a same issue regarding cutoff FPKM value and didn't find any strong basis for the same.:( As I have done with the pooled samples I am not sure about the p-values even so I am having so much trouble during data sorting. please help me out

[caggtaagtat]

Wouldn't it be better to use TPM instead of FPKM, to put the expression in the same proportion across samples?