Calculating the bisulfite conversion rate

Question

[igor]

I am trying to calculate the bisulfite conversion rates in WGBS/RRBS experiments. Bismark generates a nice report like this:

Final Cytosine Methylation Report
=================================
Total number of C's analysed:    86084529
Total methylated C's in CpG context:    9308262
Total methylated C's in CHG context:    84996
Total methylated C's in CHH context:    131989

Total C to T conversions in CpG context:    5029566
Total C to T conversions in CHG context:    19274244
Total C to T conversions in CHH context:    52255472

C methylated in CpG context:    64.9%
C methylated in CHG context:    0.4%
C methylated in CHH context:    0.3%

This seems to be the numbers that I need to use, but I am not sure how I get the bisulfite conversion rate from them. Is it possible from a single sample or do I need to have both bisulfite treated and untreated samples?

Answer 1

[Jautis]

From this information, it is not possible to say and additional/untreated samples would not help. To estimate bisulfite conversion rate, you need to have DNA that you know is unmethylated added to your sample before bisulfite conversion (for example, our lab adds 1ng lambda phage DNA to ~200ng of our mammalian samples). In that sample, your bisulfite conversion rate is the proportion of Cs converted to Ts (since all Cs are unmethylated and should be converted). This can't be determined from your experimental sample because, even for CHG and CHH sites, the actual methylation status is unknown.

[Jautis]

Moving on: Do you know if you have unmethylated DNA in the sample (and, if so, what type)? If you do, I can help you generate a bisfulfite conversion rate from that.

[igor]

Do you mean if I have a separate sample that did not undergo bisulfite conversion? I do not.

[Jautis]

No. I'm asking if you have DNA of known methylation in with your sample. Usually this is some bacterial or phage DNA.

The process should look like: (sample DNA + unmethylated DNA) --> bisulfite conversion --> sequencing --> mapping --> pull out reads mapping to your unmethylated DNA sample --> calculate conversion rate (converted/(converted + unconverted))

[igor]

Thanks for all your advice. Which Cs should be used for conversion rate? CpG, CHG, or CHH context? Or all three? Should I only be looking at the lambda phage reads to determine the rates?

[Jautis]

Only look at the lambda phage DNA to determine the rates (you don't know the methylation of your sample!). You can use all types; they should all be unmethylated regardless of the context.

[igor]

So let's say I add lambda phage to my sample, then do bisulfite conversion, then alignment to lambda phage genome, then get a certain CpG methylation rate.

For example, that rate is 3%. Does that mean bisulfite conversion rate is 97%? Is it that simple?

[Jautis]

Precisely! Just as a sanity check, in our lab it is usually over 99.5%

[hxlei613]

Does lambda phage DNA affect mapping efficiency when the mixed reads ( phage DNA & sample DNA ) are aligned to reference genome ? Thank you.

[Kangli Wang]

Hello, Jautis. Could you tell me how to extract the lambda DNA after mapping? Thanks.

[igor]

Can you clarify how you are adding lamda phage? Would you have it as a separate low concentration library? When you align it, it would be to a different genome, so doesn't that also introduce another variable?

[Jautis]

We add it to the sample so you have <1% lambda phage DNA. As such, it is in the same library (same barcodes, same conversion tube, same PCR, etc.). Yes, it will align to a different genome, but you can map it separately or together. You just want to know what percentage of sites are converted (theoretically they all should be)

Mapping to another genome doesn't introduce new variables; you're not trying to measure actual methylation in the phage DNA; just determine how well the conversion worked.