How to compute average methylation among multiple sites given raw sequencing reads

Question

[zhourudy2]

I have a question regarding calculating average methylation across all sites from one single gene. I have the raw data from chip bis sequencing in the following form.

Total reads, # of C reads, ratio
15, 5,0.33
20,5,0.4
10,5,05

My question is since the total number of reads at different sites is different, which would be a better way to compute average methylation?

1. Take the ratio and divide by number of site?
2. Take the sum of #of C reads among all sites/ sum of Total reads?

I am leaning towards method #2 since I feel like its a more accurate depiction of the raw data but I wanna hear some opinions!

Thanks!

[feljpants]

there is a typo - the second column should be 20,5,0.25

Answer 1

[Devon Ryan]

It's probably best to just take the weighted average. So for this "gene", you'd get:

0.33*(15/(15+20+10)) + 0.4*(20/(15+20+10)) + 0.5*(10/(15+20+10))

which is ~0.399.

Answer 2

[feljpants]

The simple average and the weighted average output the same methylation percentage

18/45     =        0.33*(15/(15+20+10)) + 0.4*(20/(15+20+10)) + 0.5*(10/(15+20+10))