Hello

Merging replicates is not a reasonable option. If you can do a PCA analysis, it would show all within and between variations. To present your methylation data for each DMR, you may simply use means +&- Sd.

Good luck

Different individual may have different methylation baseline. The validation of the results using targeted gene methylation analysis is recommended.

From you description it would appear you have cell culture materials? While using PCA to decide which samples to average, what you gain in reduced variation you lose in sample size. As you have such a small study this is always going to lead to 'preliminary data' and you will have to do targeted methylation analysis to provide any certainty

If you had 2 conditions, and 6 per group, you might have a better chance (again assuming its a cell culture experiment and a robust intervention)

J

Quick and cost effectively, Line-1 methylation analysis is another method to see if there is difference among two conditions. Line-1 methylation has been used in drug screening, dosing response, disease vs normal, treated vs untreated, etc.

Yes, sorry I forgot to write that the samples come from a cell line.

The idea is to match these data with an RNAseq data on same samples, to find DE-DM genes and try to validate them combining in vitro-in vivo approaches, as well as using existing databases.

Just a quick question: I have done PCA but my samples are more or less spread all over the plot. Reasoning about this with the person that performed the 850 chip, we came to the conclusion that if the samples are very similar to each other, the variation given just by different chips or lanes on the chips can give such a difference. Do you think that this hypothesis can be reliable (in your experience) ?