Question about DMR pair results

[AlvinTam-lab]

Hi,

I ran modkit dmr pair without replicates. Is it normal that no balance_map_pvalue and balance effect_size in my result bed file? Can we expect that the map_p_value will generally reduce when more replicates are included? And do you have any suggestion on the threshold for filtering map_pvalue (both the balance and non-balance)? like adjusted p<0.05. Thanks a lot.

Best,
Alvin

[ArtRand]

Hello @AlvinTam-lab,

Correct, with no replicates there won't be "balanced" metrics (they would be the same as the other metrics).

Can we expect that the map_p_value will generally reduce when more replicates are included?

Maybe, if you add more replicates it has the effect of increasing coverage. As coverage increases, small effect sizes can have lower MAP-based p-values. Once the coverage gets very high (>50X or so), the test becomes overpowered and it's more appropriate to look at effect sizes and Cohen's h values.

And do you have any suggestion on the threshold for filtering map_pvalue (both the balance and non-balance)? like adjusted p<0.05.

I would need more specifics of your study to make a useful suggestion. I generally suggest looking at uncorrected p-values < 0.01 and using Benjamini-Hochberg for correction.