--significance-factor for dmr pair in region

[hannan666666]

Hi, I'm using this code to identify DMRs (differentially methylated regions) in my target genomic regions with the '--significance-factor 0.05' parameter. I'd like to verify whether all output regions meet the significance threshold of P<0.05. Could the output include specific p-values for each region? I need these values to perform FDR correction on the results.

/home/hannan/.cargo/bin/modkit dmr pair
-a ${mkd}/${c3}/${c3}.6mA_5mC5hmC.pileup.bed.gz
-b ${mkd}/${w3}/${w3}.6mA_5mC5hmC.pileup.bed.gz
-o ${dfd}/pair_region/C_dmrpair${c3}vs${w3}cCREcpgicpgssgtf_p005.bed
--base C
--header
--ref ${ref}
--significance-factor 0.05
--regions-bed ${regions}
-f
-t 30 --log-filepath ${dfd}/pair_region/C_dmrpair${c3}vs${w3}cCREcpgicpgssgtf_p005.log

Thank you for your time !!!

[ArtRand]

Hello @hannan666666,

The --significance-factor is perhaps a misnomer, it's a parameter that the segmentation hidden Markov model uses in the emissions probabilities, and if the number is smaller the model will require more divergence between the two samples' distributions of base modifications.

Calculating the significance between regions ends up being difficult to do in a meaningful way since regions identified as different can have many CpGs. So a lot of tests will give you p-values close to zero even when the effect size is very small. For this reason I added Cohen's h statistic to the output. What I would do is sort the "different" regions by the 95%-CI lower bound.

[hannan666666]

Thank you for your reply and your updated modikit !