Why FPKM value of gene expression profiles are very close to zero?

Question

[jack]

Hi all,

I have gene expression profile of human in FPKM scale, when I take look into expression distribution, the mean is close to zero. Does it make sense ? the FPKM value for most of the genes are small and it's close to zero.

When I create box plot of expressed genes, the median is very close to zero. Is it normal or something might going wrong with my data?

Answer 1

[TriS]

if you have non-transformed FPKM/RPKM data and you plot a histogram, you will find a tall peak close to zero and then a long tail to much higher values. generally you log2 transform your data to "fix" this skewed distribution, or a few other options available with various pros-cons. this kinda of reflects what happens in a cell, you will have most of the transcript at low concentrations and a few (i.e. housekeeping) at higher concentrations/reads. however, for downstream analysis, this distribution is not the best one since it puts much weight on the low-count transcripts.

[Carlo Yague]

...this kinda of reflects what happens in a cell, you will have most of the transcript at low concentrations and a few (i.e. housekeeping) at higher concentrations/reads.

I agree. To emphasis this point, think that in a total RNA extract, you'll have about 95% rRNA (only a few genes !), the remaining 5% being split mostly between tRNAs and mRNAs.

Log transformation of your data (you might need pseudocounts) before the boxplot is a good idea.