what dataType to go with assay=CHiP-seq?

[kdaily]

We looked at ENCODE - they have what we would call assayTarget = histone and assayTarget = transcription factor. We currently have assayTarget = H3K9blah. The question remaining is which is more useful?

[kdaily]

Of note in PEC we currently have histone and transcription factor targets (CTCF).

[kdaily]

@amapeters and @sgosline please weigh in!

[sgosline]

Wait is this a question about assayTarget or dataType?

[sgosline]

i think the specific histone mark is helpful, I mean k27me3 and k4me3 are diametrically opposed to one another. But I don't have any of this data in my communities so I'm purely arguing from a scientific standpoint.

[sgosline]

If we're talking about dataType then i think chromatinActivity is appropriate

[amapeters]

We use assayTarget to specify which histone mark, TF, or input

[amapeters]

If we use assayTarget - histone, we need an additional level to specify which mark. I prefer it the way it is now

[sgosline]

I agree with @amapeters

[kdaily]

@sgosline the question @allaway raised is that chromatinActivity is not an appropriate value for ChIP-Seq applied to transcription factors, for example.

[sgosline]

I disagree - regardless of how you believe TFs to be altering the confirmation of the chromatin alongside histones and molecules like CTCF (which we once called a TF), they are binding to chromatin and altering the activity around them.

[amapeters]

Is 'chromatinState' more accurate. Or, we just call it 'epigenomics'?

[sgosline]

chromatinState seems so similar it's not worth the change, though I don't mind it if it makes things easier.

Epigenomics is still a somewhat contested term (I was not allowed to submit it in the title of my 2015 paper for example) last time I checked and can refer to any non-transcriptional regulatory mechanisms such as NMD, miRNA regulation, etc.

[amapeters]

chromatinState would also be appropriate for ATACseq (i.e., probing open chromatin)

[sgosline]

Sure!

[kdaily]

Not all TFs are chromatin associated, as CTCF is though?

[sgosline]

I feel like I'm missing the issue here - If you are performing ChIP-Seq you are by definition assaying the chromatin state/activity, regardless of how the TF is directly interacting with it and/or affecting it.

[kdaily]

Very true! I thought the chromatin in chromatinActivity was introduced to described the target - all of our ChIP-Seq data is histone/chromatin modification related.

If I was coming into a dataset and was looking for TF-binding datasets, knowing that ChIP-Seq is a common technology for that, my first inclination would not be to use chromatinActivity as a way to find that.

Basically comes back to our 'technical description' versus 'intent' issues with other terms.

[sgosline]

I'm not entirely sure, if you read through the ENCODE papers it becomes clear that the model of a transcription factor merely activating or repressing transcription is over simplified. So if your intent is to uncover the complete role of TFs in gene regulation you need a lot more than ChIP-Seq data...

[kdaily]

ENCODE still thinks it's useful (?) to differentiate in their data between histone and non-histone ChIP-Seq data, per the way they annotate.

[sgosline]

If I read correctly, so do we, by specifying the assayTarget.

[amapeters]

Yes, I thought we decided that already. assay - ChIPseq, assayTarget - specific histone mark or TF

[amapeters]

• and input, when no Ab

[kdaily]

Their filter is Assay Category = 'DNA binding', then Target of Assay = ['histone', 'transcription factor', ...].

https://www.encodeproject.org/matrix/?type=Experiment&status=released&assay_slims=DNA+binding&target.investigated_as=histone

[kdaily]

The question is (as @sgosline is getting at) if there is a meaningful distinction between datasets interrogating histone modifications and those that aren't. With the current solution, If I want to find all of either one exclusively, I'm going to have to enumerate histone marks or TF names.

ENCODE seemed to think that it's something users want to be able to do. Not that ENCODE is always right, but we have looked to them as a model in the past.

[sgosline]

Not in complete agreement but can re-open when we have more specific use case for transcription factor searches. Decision is to use chromatinActivity