Interpretation of Beta values : Methylation data

Question

[Tanvir Ahamed]

Beta values (β) are the estimate of methylation level using the ratio of intensities between methylated and unmethylated alleles. β are between 0 and 1 with 0 being unmethylated and 1 fully methylated.

Now when compare values of a probe form 2 different sample, how to compare if methylated or unmethylated . As for example,

probe  sample_A  Sample_B
cg_1      0.6       0.7
cg_2      0.2       0.3
cg_3      0.8       0.9
cg_4      0.2       0.9
cg_5      0.3       0.6
cg_6      0.1       0.4

Now how to compare a specific probe if methylated or not.

Some assumption :

1. If a specific value ( 0.5) define methylated (greater than 0.5) or unmethylated (less than 0.5) state.
2. If it is reasonable to measure the change ( difference) of beta values. For cg_6 difference is 0.3 (0.4-0.1), though the maximum value is less than 0.5.
3. Or if consider both situation together.

May be the issue is really simple. But i am a bit unclear regarding this issue. I am expecting some expert's thinking on this issue.

Partly related : Interpreting Fractional Methylation Data

Thanks.

Answer 1

[Devon Ryan]

1. This assumption is incorrect. A position is only ever has a binary methylated/unmethylated state on a single copy of a chromosome in a single cell. A value of 0.5 typically indicates that the underlying cells that were sampled are highly variable (it's the unclear whether there are multiple 0% and 100% methylation populations or if there's a more heterogeneous mix).
2. Yes, this is reasonable, though in reality there's noise around the estimate of 0.1 and 0.4 and this should be taken into consideration.
3. This will depend completely on what you want to do.

There are already a number of Bioconductor packages for handling methylation data. I would strongly encourage you to use one of them and not try to come up with your own methods.

[Tanvir Ahamed]

Thanks for your thought. Can you suggest some function/package those handle this type of situation ?

[Devon Ryan]

It depends on the exact question and the dataset. In general, just look through the Bioconductor DNAMethylation view and you'll be able to narrow down the options to a couple possibilities.

[vinayjrao]

Dear Devon,

In this case (1), is it OK to consider that a value of 0.1 would indicate that 10% of my samples were methylated, while the rest were not?

Thanks in advance

[Chip]

Not 10% percent of your samples but 10% of cells in your sample for that site.