All Comb1-n "NA" for most DMRs in CHH

[mkpython3]

Hey, i am using HOME-timeseries with 23 Samples (no replicates) and i just noticed that most of my DMRs that got reported by HOME have "NA" in every Comb1-n column, even though max_delta and confidence_scores are reported.

An example looks like this:

chr	start	end	numC	len	max_delta	confidence_scores	HvBS001_VS_HvBS002	HvBS001_VS_HvBS004	HvBS001_VS_HvBS005	HvBS001_VS_HvBS006	HvBS001_VS_HvBS007	HvBS001_VS_HvBS008	HvBS001_VS_HvBS009	HvBS001_VS_HvBS011	HvBS001_VS_HvBS012	HvBS001_VS_HvBS013	HvBS001_VS_HvBS014	HvBS001_VS_HvBS015	HvBS001_VS_HvBS016	HvBS001_VS_HvBS017	HvBS001_VS_HvBS018	HvBS001_VS_HvBS019	HvBS001_VS_HvBS020	HvBS001_VS_HvBS021	HvBS001_VS_HvBS022	HvBS001_VS_HvBS023	HvBS001_VS_HvBS024	HvBS001_VS_HvBS048	HvBS002_VS_HvBS004	HvBS002_VS_HvBS005	HvBS002_VS_HvBS006	HvBS002_VS_HvBS007	HvBS002_VS_HvBS008	HvBS002_VS_HvBS009	HvBS002_VS_HvBS011	HvBS002_VS_HvBS012	HvBS002_VS_HvBS013	HvBS002_VS_HvBS014	HvBS002_VS_HvBS015	HvBS002_VS_HvBS016	HvBS002_VS_HvBS017	HvBS002_VS_HvBS018	HvBS002_VS_HvBS019	HvBS002_VS_HvBS020	HvBS002_VS_HvBS021	HvBS002_VS_HvBS022	HvBS002_VS_HvBS023	HvBS002_VS_HvBS024	HvBS002_VS_HvBS048	HvBS004_VS_HvBS005	HvBS004_VS_HvBS006	HvBS004_VS_HvBS007	HvBS004_VS_HvBS008	HvBS004_VS_HvBS009	HvBS004_VS_HvBS011	HvBS004_VS_HvBS012	HvBS004_VS_HvBS013	HvBS004_VS_HvBS014	HvBS004_VS_HvBS015	HvBS004_VS_HvBS016	HvBS004_VS_HvBS017	HvBS004_VS_HvBS018	HvBS004_VS_HvBS019	HvBS004_VS_HvBS020	HvBS004_VS_HvBS021	HvBS004_VS_HvBS022	HvBS004_VS_HvBS023	HvBS004_VS_HvBS024	HvBS004_VS_HvBS048	HvBS005_VS_HvBS006	HvBS005_VS_HvBS007	HvBS005_VS_HvBS008	HvBS005_VS_HvBS009	HvBS005_VS_HvBS011	HvBS005_VS_HvBS012	HvBS005_VS_HvBS013	HvBS005_VS_HvBS014	HvBS005_VS_HvBS015	HvBS005_VS_HvBS016	HvBS005_VS_HvBS017	HvBS005_VS_HvBS018	HvBS005_VS_HvBS019	HvBS005_VS_HvBS020	HvBS005_VS_HvBS021	HvBS005_VS_HvBS022	HvBS005_VS_HvBS023	HvBS005_VS_HvBS024	HvBS005_VS_HvBS048	HvBS006_VS_HvBS007	HvBS006_VS_HvBS008	HvBS006_VS_HvBS009	HvBS006_VS_HvBS011	HvBS006_VS_HvBS012	HvBS006_VS_HvBS013	HvBS006_VS_HvBS014	HvBS006_VS_HvBS015	HvBS006_VS_HvBS016	HvBS006_VS_HvBS017	HvBS006_VS_HvBS018	HvBS006_VS_HvBS019	HvBS006_VS_HvBS020	HvBS006_VS_HvBS021	HvBS006_VS_HvBS022	HvBS006_VS_HvBS023	HvBS006_VS_HvBS024	HvBS006_VS_HvBS048	HvBS007_VS_HvBS008	HvBS007_VS_HvBS009	HvBS007_VS_HvBS011	HvBS007_VS_HvBS012	HvBS007_VS_HvBS013	HvBS007_VS_HvBS014	HvBS007_VS_HvBS015	HvBS007_VS_HvBS016	HvBS007_VS_HvBS017	HvBS007_VS_HvBS018	HvBS007_VS_HvBS019	HvBS007_VS_HvBS020	HvBS007_VS_HvBS021	HvBS007_VS_HvBS022	HvBS007_VS_HvBS023	HvBS007_VS_HvBS024	HvBS007_VS_HvBS048	HvBS008_VS_HvBS009	HvBS008_VS_HvBS011	HvBS008_VS_HvBS012	HvBS008_VS_HvBS013	HvBS008_VS_HvBS014	HvBS008_VS_HvBS015	HvBS008_VS_HvBS016	HvBS008_VS_HvBS017	HvBS008_VS_HvBS018	HvBS008_VS_HvBS019	HvBS008_VS_HvBS020	HvBS008_VS_HvBS021	HvBS008_VS_HvBS022	HvBS008_VS_HvBS023	HvBS008_VS_HvBS024	HvBS008_VS_HvBS048	HvBS009_VS_HvBS011	HvBS009_VS_HvBS012	HvBS009_VS_HvBS013	HvBS009_VS_HvBS014	HvBS009_VS_HvBS015	HvBS009_VS_HvBS016	HvBS009_VS_HvBS017	HvBS009_VS_HvBS018	HvBS009_VS_HvBS019	HvBS009_VS_HvBS020	HvBS009_VS_HvBS021	HvBS009_VS_HvBS022	HvBS009_VS_HvBS023	HvBS009_VS_HvBS024	HvBS009_VS_HvBS048	HvBS011_VS_HvBS012	HvBS011_VS_HvBS013	HvBS011_VS_HvBS014	HvBS011_VS_HvBS015	HvBS011_VS_HvBS016	HvBS011_VS_HvBS017	HvBS011_VS_HvBS018	HvBS011_VS_HvBS019	HvBS011_VS_HvBS020	HvBS011_VS_HvBS021	HvBS011_VS_HvBS022	HvBS011_VS_HvBS023	HvBS011_VS_HvBS024	HvBS011_VS_HvBS048	HvBS012_VS_HvBS013	HvBS012_VS_HvBS014	HvBS012_VS_HvBS015	HvBS012_VS_HvBS016	HvBS012_VS_HvBS017	HvBS012_VS_HvBS018	HvBS012_VS_HvBS019	HvBS012_VS_HvBS020	HvBS012_VS_HvBS021	HvBS012_VS_HvBS022	HvBS012_VS_HvBS023	HvBS012_VS_HvBS024	HvBS012_VS_HvBS048	HvBS013_VS_HvBS014	HvBS013_VS_HvBS015	HvBS013_VS_HvBS016	HvBS013_VS_HvBS017	HvBS013_VS_HvBS018	HvBS013_VS_HvBS019	HvBS013_VS_HvBS020	HvBS013_VS_HvBS021	HvBS013_VS_HvBS022	HvBS013_VS_HvBS023	HvBS013_VS_HvBS024	HvBS013_VS_HvBS048	HvBS014_VS_HvBS015	HvBS014_VS_HvBS016	HvBS014_VS_HvBS017	HvBS014_VS_HvBS018	HvBS014_VS_HvBS019	HvBS014_VS_HvBS020	HvBS014_VS_HvBS021	HvBS014_VS_HvBS022	HvBS014_VS_HvBS023	HvBS014_VS_HvBS024	HvBS014_VS_HvBS048	HvBS015_VS_HvBS016	HvBS015_VS_HvBS017	HvBS015_VS_HvBS018	HvBS015_VS_HvBS019	HvBS015_VS_HvBS020	HvBS015_VS_HvBS021	HvBS015_VS_HvBS022	HvBS015_VS_HvBS023	HvBS015_VS_HvBS024	HvBS015_VS_HvBS048	HvBS016_VS_HvBS017	HvBS016_VS_HvBS018	HvBS016_VS_HvBS019	HvBS016_VS_HvBS020	HvBS016_VS_HvBS021	HvBS016_VS_HvBS022	HvBS016_VS_HvBS023	HvBS016_VS_HvBS024	HvBS016_VS_HvBS048	HvBS017_VS_HvBS018	HvBS017_VS_HvBS019	HvBS017_VS_HvBS020	HvBS017_VS_HvBS021	HvBS017_VS_HvBS022	HvBS017_VS_HvBS023	HvBS017_VS_HvBS024	HvBS017_VS_HvBS048	HvBS018_VS_HvBS019	HvBS018_VS_HvBS020	HvBS018_VS_HvBS021	HvBS018_VS_HvBS022	HvBS018_VS_HvBS023	HvBS018_VS_HvBS024	HvBS018_VS_HvBS048	HvBS019_VS_HvBS020	HvBS019_VS_HvBS021	HvBS019_VS_HvBS022	HvBS019_VS_HvBS023	HvBS019_VS_HvBS024	HvBS019_VS_HvBS048	HvBS020_VS_HvBS021	HvBS020_VS_HvBS022	HvBS020_VS_HvBS023	HvBS020_VS_HvBS024	HvBS020_VS_HvBS048	HvBS021_VS_HvBS022	HvBS021_VS_HvBS023	HvBS021_VS_HvBS024	HvBS021_VS_HvBS048	HvBS022_VS_HvBS023	HvBS022_VS_HvBS024	HvBS022_VS_HvBS048	HvBS023_VS_HvBS024	HvBS023_VS_HvBS048	HvBS024_VS_HvBS048
chr1H	26732	26836	7	104	0.16683316683316682	0.25653578517793324	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA

Is this expected behavior? How can this happen?

I am happy to provide more information if necessary. In the meantime i will check if this happened also for other Contexts.
The input files are about 273M (uncompressed), so let me know how to submit them if they are needed.
Thanks alot in advance.

[Akanksha2511]

Hi,

Sorry for the delay in reply. NA implies that the status (hyper/hypo) of the DMRs could not be interpreted. I would suggest visualisation of few DMRs. But, if all the status is "NA" I would say it does not have enough power to identify the DMRs with confidence.

Thanks,
Akanksha

[mkpython3]

Hey, thank you for your response.

So do I understand it correctly: HOME first discovers a potential DMR in that region. But then it does not find enough confidence to report it. The DMR will be written to the output file anyways but with all columns NA.
I just think it is a little weird since it only happened in the CHH context and a confidence score is still reported. Maybe the DMR should not appear in the output at all then?

Best regards
Marius

[Akanksha2511]

Hi Marius,

Yes, so for time series data, HOME will try to find the difference between all the samples and will report the region of difference and its confidence (delta). It then checks for a consistent delta but is same direction (hyper/hypo) for possible pairwise comparisons. If its not able to find that it will report NA. You are right its a bit weird to get 'NAs' in all the samples. I guess its just telling you that the data does not have enough coverage for CHH methylation identification. Have to tried to visualise these DMRs on IGV or UCSC browser?

Thanks,
Akanksha

[mkpython3]

Hey Akanksha

I have visualized a few DMRs now using a Python script. For some of the "All NA" DMRs it is quite obvious why they are NA because there is rarely any data in the range of the DMR. But for others i think there should be enough data that coverage should not be an issue. I have attached a few plots below. Here the size of the dots represents the coverage and the color (blue to pink) the level of methylation.

These are the plots of DMRs that have >=50% of data in the Comb1-N columns:

These are DMRs where there is 100% NA in every column like in the first post I made, where I think it should be possible to call a DMR:

This is one of the 100% NA DMRs where I think its quite obvious why there is so much NA, but I just think it maybe should not be outputted at all:

Thanks alot for your help!
Marius

[Akanksha2511]

Hi Marius,

Thanks for sharing the plots. Its hard to visualise it like this. I think it will better to have a IGV or a UCSC plot in the standard format which will also show the direction of methylation. Please have a look at the HOME paper for reference.

Thanks,
Akanksha

[mkpython3]

Hey Akanksha,

I managed to visualize the methylation levels in IGV. However, I don't know how to reproduce the plots with the methylation difference like in the HOME paper. And btw, what do you mean with "direction of methylation". Do you include an extra track with the DMR information? Because the methylation files are only 0 - 1 and not -1 - 1. Or is there an option in IGV to produce a methylation difference track from the others?

Best regards,
Marius

[Akanksha2511]

Hi Marius,

You can upload the wig files from BSseeker2 for the samples and visualise the DMRs on IGV. I think this should give us the idea about what's the issue.

Thanks,
Akanksha

[mkpython3]

Hey Akanksha,

thank you for your response. I am using Bismark for my analysis and merged the + and - strand for all chromosomes, therefore i have bedgraph files that go from 0 to 1 instead of wig files that go from -1 to 1. Anyways it should not make a huge difference. I converted those bedgraph files to .tdf files for igv and made a screenshot of a 100% NA DMR. I hope this helps?

Best regards,
Marius