Dear all,
Can I get some enlightenment from any of you who has more experience than I do to really understand a transcriptomic analysis result? How to determine that one gene is less expressed and the other is more expressed (judging from rpkm or fpkm value)? Thank you for any input.
first you need to define a reference gene, usually it's a house keeping gene such as ACTB, GAPDH, or RNase-coding RNAs, level of your reference is pre-defined as uniform among different samples, and relative expression level as compared to your reference is commonly used to define expression level.
You have to perform a formal Differential Expression Analysis. If you already have the data counts (rpkm or fpkm ) for your 2 groups (treatment 1 vs treatment 2), you can run it in R with a package such as edgeR or DEseq. In that way you will get a list of the genome-wide deferentially expressed genes.
If you only have FPKM data, I don't know of any good statistical measures to conclusively say that one gene is more highly expressed than another because these measures are already normalized for read depth and so that statistically crucial information about total read depth is lost. IMPORTANT: FPKM/RPKM are not the same as counts and you can not use them as inputs for DEseq/edgeR/EBseq etc. If you only have FPKM data, Chen is right, relative expression is really the best you can do. This will be more Exploratory Data Analysis than statistical proof, though.
Where did your data come from? Ideally you also have counts/expected counts from RSEM, Kallisto, the tuxedo suite etc to which you can apply a differential expression analysis method.
You won't be able to determine it statistically without counts and replicates. The FPKM values are still a real measure of the proportional number of transcripts detected in each sample, so if a gene has FPKM higher values in sample A compared to sample B, this is [inconclusive] evidence that the gene is expressed at higher levels in sample A (as long as the samples have the same makeup of cell types, are the same species etc). Make sure you're thinking about what is being measured, and the biology behind it. This is a good guide of what RNAseq units mean:
https://haroldpimentel.wordpress.com/2014/05/08/what-the-fpkm-a-review-rna-seq-expression-units/
Dear Angga,
like Erik already mentioned, I can only recommend the R-package DeSeq2. There's a really extensive and helpful tutorial "Analyzing RNA-seq data with DESeq2" by Michael I. Love, Simon Anders, and Wolfgang Huber:
https://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html
Sorry for the late answer, just stumbled across your post right now.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Values