About FPKM of interested genes

Question

[Ginsea Chen]

Hi ALL

Today I tried to use FPKM to obtain the expression characteristic of my interested gene in specified tissue based on RNA-seq.

My operation is follows:

1. Converted sra (download from NCBI) file to fastq file and then converted it to fasta file(RNA-seq data);
2. Blast was used to find fragments which have zero E-value to interested gene
3. Extracted information of these fragments trough fastq file and then converted it as fastq format
4. Bowtie2 and cufflinks were used to calculate the FPKM values associated with reference genome

In the out file of cufflinks, such as gene.fpkm_tracking and others, the FPKM values were not only one. So I don't know how to use these FPKM values indicate interested gene expression, average value or total value ?

Thanks all.

Answer 1

[Charles Warden]

I would recommend using the gene.fpkm_tracking files. I think the gene-level quantification is more reliable than the isoform predictions (I tend to favor splicing event analysis over whole-isoform differential expression because I haven't seen any alignment that I thought supported the prediction of one isoform truly showing very different expression than the other isoforms).

Within that file, here is a description of the columns.

If you want help extracting the FPKM values, there is an RNA.prepare.inpu() in the sRAP package

You can also use cuffdiff to identify differential expressed genes without needing to interpret the current files, but I wouldn't recommend that as a preferred method because think it can sometimes give weird results. For example, here is a comparison of differential expression analysis.

[Ginsea Chen]

Thanks for your help, best regards.