Question about DMR pair results

[zine-el-aabidine]

Dear,

Thank you for your efforts in updating the version following the requested changes.

I ran DMR pair twice, with and without the --min-valid-coverage 10 option. For the same modification site, I obtained different a_count and b_count values, as well as different statistics. I ran the same command on the same file using --base A and --single-code "a".

I would be very grateful for your help.

Kind regards,

<style> </style>

	#chrom	start	end	name	score	strand	a_counts	a_total	b_counts	b_total	a_mod_percentages	b_mod_percentages	a_pct_modified	b_pct_modified	map_pvalue	effect_size	balanced_map_pvalue	balanced_effect_size	pct_a_samples	pct_b_samples	replicate_map_pvalues	replicate_effect_sizes	cohen_h	cohen_h_low	cohen_h_high
without_min_coverage	2L	21012	21013	.	1.9818691992013413	-	a:0	31	a:3	28	a:0.00	a:10.71	0	0.10714286	0.4706322371851205	-0.10714285714285714	1	-0.037037037037037035	100	100	0.45672346927925245,1,1	-0.2,0,0	-0.66694634794494	0.155954191101083	1.1779385047887971
with min_coverage_10	2L	21012	21013	.	2.729244391195497	-	a:0	24	a:3	15	a:0.00	a:20.00	0	0.2	0.1263947920523823	-0.2	0.1263947920523823	-0.2	66	33	-	-	-0.9272952254521929	0.2821921027589833	1.5723983481454025

[zine-el-aabidine]

Hi again,
I think this difference is due to filtering of replicates that have fewer reads than the threshold. The problem is that I would like to apply a threshold of 10 reads for condition B and 5 reads for condition A, but I’m not sure how to do this.
Thank you again

[ArtRand]

Hello @zine-el-aabidine,

I see, that's not an option right now. I could add it if you think there this is a common use case. You could still do what you want by pre-filtering the bedmethyl files on valid coverage (5 for A and 10 for B).

[zine-el-aabidine]

Hi @ArtRand
Thank you for your response. In our model, methylation is blocked in the mutant, and our goal is to identify the most reliable m6A methylation sites in the WT by comparing them with the mutant. The idea is to perform a DMR analysis while lowering the threshold for valid coverage in the mutant samples.

We observed a large number of sites that pass the threshold of 10 valid reads in all three WT replicates but show no valid coverage in the mutant when applying the same threshold. Therefore, we would like to adjust the coverage threshold and evaluate how this affects the results.

I will try the option you suggested. The bedMethyl files, i have, contain both types of adenine base modifications.

Best

[ArtRand]

Hello @zine-el-aabidine,

The idea is to perform a DMR analysis while lowering the threshold for valid coverage in the mutant samples.

Do you mean that you want to lower the threshold of m6A reads? I don't understand why you want to lower the coverage from the mutant sample, with less coverage you'll have more uncertainty about the difference in methylation levels.

We observed a large number of sites that pass the threshold of 10 valid reads in all three WT replicates but show no valid coverage in the mutant when applying the same threshold. Therefore, we would like to adjust the coverage threshold and evaluate how this affects the results.

Interesting. Could you (1) tell me what the estimated A threshold value is for the mutant sample and (2) see what happens when you pass the --no-filtering flag to pileup?