how to calculate RPKM from mean per-base read coverage

[linp5]

Hi all:

gene expression level can be presented by mean per-base read coverage(Rz).
It was reported by some researchers that they can calculate RPKM from this mean per-base read coverage.

Is there anyone can explain the details of this calculation to me ?
Thanks in advance!
Best,

[dpryan]

You should really be able to figure this out. Multiply Rz by the transcript/gene/whatever length and divide by the average read/fragment length and you have the number of reads/fragments per feature. You can then get RPKM.

[linp5]

Thanks Dpryan!

Now I calculate RPKM as below according to[1]:

((Rz * Length_of_the_transcript) / read_length) * 10^9
/
(Length_of_the_transcript * total_reads_in_the_sample)

I am still not sure about the validity of the calculation
and how to find the total number of reads?

Could you please give me some more tips ?


[1] https://tcga-data.nci.nih.gov/tcgafi...ESCRIPTION.txt
Step 4: Calculate quantification at the transcript level.

[blakeoft]

The number of reads is the ((Rz * Length_of_the_transcript) / read_length) part.

I'll try and explain.

Let N be the number of bases being considered, i.e. the length of the transcript. Let c_i be the coverage of base i for i in 1, 2, ..., N, and let L be the length (or average length) of the reads. You can write the above expression as (((1/N times the sum from i = 1 to N of c_i) * N) / L). The N and 1/N multiply to 1. So then we have ((sum from i = 1 to N of c_i) / L). The sum of the c_i's tells you how many bases appear among all of the reads, so dividing this sum by L will tell you how many reads were mapped to this transcript.