Re: Track hubs

Grover Aayush

unread,

Jun 26, 2025, 12:31:23 PMJun 26

to Maximilian Haeussler, Valentina Boeva, gen...@soe.ucsc.edu

Dear Maximilian,

Thank you for your interest in hosting UniversalEPI’s predictions on UCSC Genome Browser. We have exported the model predictions corresponding to all bulk ATAC-seq data from ENCODE (consisting of 157 cell lines and primary cells).

The hub can be found here: https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/ucsc_track_hub/hub.txt

Information on the input ATAC-seq tracks and the generated Hi-C can be found here: https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/

It would be wonderful if this hub can be to the UCSC’s list of public track hubs. Please let us know if there are any questions or concerns.

Best regards,

Aayush Grover
Computational Cancer Genomics Lab, ETH Zürich

Institute for Machine Learning, Department of Computer Science

CAB G 31.1

Universitätstrasse 6

8092 Zürich, Switzerland

From: Boeva Valentina <valenti...@inf.ethz.ch>
Date: Friday, 22 November 2024 at 10:22
To: Maximilian Haeussler <mhae...@ucsc.edu>
Cc: Grover Aayush <aayush...@inf.ethz.ch>, Krymova Ekaterina <ekaterin...@sdsc.ethz.ch>, Rabuzin Lovro <lovro....@inf.ethz.ch>
Subject: RE: Track hubs

Dear Max,

We are interested😊We should have the final model in the coming weeks.

Maybe we could also have a track with predicted hi-C_QTL (if we get enough computing power) ;-) We’ll think about it!

Best wishes

Valentina

From: Maximilian Haeussler <mhae...@ucsc.edu>
Sent: Thursday, November 21, 2024 22:32
To: Boeva Valentina <valenti...@inf.ethz.ch>
Cc: Grover Aayush <aayush...@inf.ethz.ch>; Krymova Ekaterina <ekaterin...@sdsc.ethz.ch>
Subject: Re: Track hubs

Yes, or the HTAN cell lines. They’ll have a lot of hic to train on.

Sharing it via UCSC should be as easy as putting all files into a directory and running a single command. Let me know if you're interested.

On Thu, Nov 21, 2024 at 6:54 AM Boeva Valentina <valenti...@inf.ethz.ch> wrote:

Dear Max,

Thank you for your email! Indeed, we could potentially predict Hi-C for all ENCODE cell lines and share the predictions via UCSC! Great idea!

Will be in touch!

Valentina

From: Maximilian Haeussler <mhae...@ucsc.edu>
Sent: Friday, November 15, 2024 10:49
To: Boeva Valentina <valenti...@inf.ethz.ch>
Subject: Track hubs

Hi Valentina,

thanks for your fascinating talk today at CSHL GI. I work for the UCSC Genome Browser and would love to look at your predictions, once they're published.

To make that easier, it would be great if your group could make the data available in a format that can be easily browsed. Your students can make bigWig or bigBed file from your data, maybe even in bigInteract format see https://genome.ucsc.edu/goldenPath/help/bigBed.html#Ex3, they can put that into a track hub, just a little text file, https://genome.ucsc.edu/goldenpath/help/hubQuickStart.html and send its URL to gen...@soe.ucsc.edu so we can add it to our list here https://genome.ucsc.edu/cgi-bin/hgHubConnect The hub in this format can also be loaded into IGV, Ensembl, NCBI, especially IGV, so anyone who has the link or is on the track hub list, can access your tracks with a single click.

If there are any problems with data hosting or format conversions or any other questions, let me know, ideally also CC gen...@soe.ucsc.edu.

I'll try to find you later, but in case you're busy, at least I sent you this email :-)

Max

Gerardo Perez

unread,

Jul 1, 2025, 3:58:24 PMJul 1

to Grover Aayush, Valentina Boeva, gen...@soe.ucsc.edu

Hello,

Thank you for sharing your UniversalEPI predictions hub and submitting a request to add it as a public track hub.

We have added this track hub to the queue. From an initial look, it appears that the UniversalEPI predictions hub is missing the following, which are required for public hubs:

A descriptionUrl html page specified in your hub.txt. This should be a URL to a description page for your entire hub. Public hubs often link to a full-text paper or to a laboratory webpage that describes the research presented in the hub.
A description page for every configuration page (composite, superTrack, or standalone track). Note that multiple tracks and/or composites can use the same description page with the 'html' setting. Here are some track description pages from public hubs:

A contact email address prominently displayed on the description pages.

You can read more about the required and recommended guidelines on the following help page:
https://genome.ucsc.edu/goldenPath/help/publicHubGuidelines.html

Let us know if you have any questions regarding your hub or track hub settings, and when you would like us to take another look at the hub.

Please include gen...@soe.ucsc.edu in any replies to ensure visibility by the team. All messages sent to that address are archived on a publicly-accessible Google Groups forum. If your question includes sensitive data, you may send it instead to genom...@soe.ucsc.edu.

Gerardo Perez
UCSC Genomics Institute

--

---
You received this message because you are subscribed to the Google Groups "UCSC Genome Browser Public Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to genome+un...@soe.ucsc.edu.
To view this discussion visit https://groups.google.com/a/soe.ucsc.edu/d/msgid/genome/6BE8BFBC-7BEC-4A71-A4A0-99A40EF73C8F%40ethz.ch.

Grover Aayush

unread,

Jul 3, 2025, 12:07:39 PMJul 3

to Gerardo Perez, Valentina Boeva, gen...@soe.ucsc.edu

Hi Gerardo,

Thank you for pointing us to the missing URLs and sorry for missing them. The track hub is now updated with these details (https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/ucsc_track_hub/hub.txt). Please let us know if anything else needs to be added.

Many thanks,

Aayush

From: Gerardo Perez <gpe...@ucsc.edu>
Date: Tuesday, 1 July 2025 at 21:58
To: Grover Aayush <agr...@ethz.ch>
Cc: Valentina Boeva <valenti...@inf.ethz.ch>, "gen...@soe.ucsc.edu" <gen...@soe.ucsc.edu>
Subject: Re: [genome] Re: Track hubs

Hello,

Thank you for sharing your UniversalEPI predictions hub and submitting a request to add it as a public track hub.

We have added this track hub to the queue. From an initial look, it appears that the UniversalEPI predictions hub is missing the following, which are required for public hubs:

A descriptionUrl html page specified in your hub.txt. This should be a URL to a description page for your entire hub. Public hubs often link to a full-text paper or to a laboratory webpage that describes the research presented in the hub.

Gerardo Perez

unread,

Jul 15, 2025, 8:09:40 PMJul 15

to Grover Aayush, Valentina Boeva, gen...@soe.ucsc.edu

Hello,

Thank you for updating the track hub.

We took a look at the hub and noticed that the Hi-C Predictions tracks were built using the bigInteract track format. Did you consider using the hic track format to build these tracks?

We also noticed that some of the tracks in the Hi-C Predictions composite track have “raw” in their label, such as the A549 raw track. Were these tracks processed using the log-transformed ICE-normalized Hi-C format?

This is not required, but you may be interested to know that you can also build tracks using the dimensions setting in the trackDb file. This setting allows for a two-dimensional array of checkboxes based on groups in the track configuration interface:

In the trackDb file, you could add subGroup2 and dimensions settings to the universalepi_hic composite track, such as the following:

track universalepi_hic
compositeTrack on
shortLabel Hi-C Predictions
longLabel UniversalEPI Hi-C Predictions using ATAC-seq from ENCODE
type bigInteract
endsVisibility two
visibility full
html https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/ucsc_track_hub/hg38/HiC.html
subGroup1 dataType Data_Type raw=Raw z_score=Z-score
subGroup2 cellType Cell_Type HCT116=HCT116 A549=A549
dimensions dimX=dataType dimY=cellType

Then, for each subtrack of the universalepi_hic composite track, you can include the cellType in the subGroups line, for example:

track hct116
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/raw/hct116.bb
shortLabel hct116
longLabel HCT116 raw
parent universalepi_hic off
subGroups dataType=raw cellType=HCT116
visibility hide

track z_score_hct116
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/z_score/hct116_z.bb
shortLabel z_score_hct116
longLabel HCT116 z-score
parent universalepi_hic off
subGroups dataType=z_score cellType=HCT116

track a549_rgs
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/raw/a549_rgs.bb
shortLabel a549_rgs
longLabel A549 raw
parent universalepi_hic off
subGroups dataType=raw cellType=A549
visibility hide

track z_score_a549_rgs
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/z_score/a549_rgs_z.bb
shortLabel z_score_a549_rgs
longLabel A549 z-score
parent universalepi_hic off
subGroups dataType=z_score cellType=A549
visibility hide

Here is a snippet of the trackDb file with these settings:
https://hgwdev.gi.ucsc.edu/~gperez2/mlq/hub_35990/UniversalEPI/hg38/trackDb_edit.txt

Here is a session with the UniversalEPI ENCODE hub attached, and these trackDb edits applied:
https://genome.ucsc.edu/s/gperez2/UniversalEPI_dimensions

I hope this is helpful. Please include gen...@soe.ucsc.edu in any replies to ensure visibility by the team. All messages sent to that address are archived on a publicly-accessible Google Groups forum. If your question includes sensitive data, you may send it instead to genom...@soe.ucsc.edu.

Gerardo Perez
UCSC Genomics Institute

Grover Aayush

unread,

Jul 17, 2025, 3:29:38 PMJul 17

to Gerardo Perez, Valentina Boeva, gen...@soe.ucsc.edu

Dear Gerardo,

Thank you for the suggestions.

We went with bigInteract as these files are seen in the arc format by default. This is ideal in our case because we predict chromatin interactions only between the regulatory elements (and not for all the 5Kb bins in the genome). The complete hic track format for our predictions looks quite sparse and hence, not an ideal visualization format. I understand that hic files can also be visualized as arcs but I wasn’t sure if this can be set by default. That being said, if you think that the hic track format is better, we can easily generate these files.
Indeed, the “raw” tracks correspond to the log-transformed ICE-normalized Hi-C. Thank you for asking this as it was not clear in the description. We have now modified the description to clarify this.
Dimensions setting is a very neat solution for our track hub that has 150+ tracks. We have now included this in our trackDb file. Your examples very helpful.

The hub is now modified with these changes (https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/ucsc_track_hub/hub.txt). Looking forward to your comments.

Many thanks,

Aayush

From: Gerardo Perez <gpe...@ucsc.edu>
Date: Wednesday, 16 July 2025 at 02:09
To: Grover Aayush <agr...@ethz.ch>
Cc: Valentina Boeva <valenti...@inf.ethz.ch>, "gen...@soe.ucsc.edu" <gen...@soe.ucsc.edu>
Subject: Re: [genome] Re: Track hubs

Hello,

Thank you for updating the track hub.

We took a look at the hub and noticed that the Hi-C Predictions tracks were built using the bigInteract track format. Did you consider using the hic track format to build these tracks?

We also noticed that some of the tracks in the Hi-C Predictions composite track have “raw” in their label, such as the A549 raw track. Were these tracks processed using the log-transformed ICE-normalized Hi-C format?

This is not required, but you may be interested to know that you can also build tracks using the dimensions setting in the trackDb file. This setting allows for a two-dimensional array of checkboxes based on groups in the track configuration interface:

Gerardo Perez

unread,

Jul 29, 2025, 12:15:00 PMJul 29

to Grover Aayush, Valentina Boeva, gen...@soe.ucsc.edu

Hello, Aayush.

Thanks for the explanation. The current display feels very dense, with too many connections. It might help to apply stronger filters, use the hic format, or try a filtered version with hic. Let us know if you would like to explore that option.

We did notice that some tracks share the same longLabel, which can be confusing when displayed in the browser graphic.

https://genome.ucsc.edu/s/gperez2/universalepi_hic

For example, the a549_rgs and a549_asc tracks have the same longLabel:

track a549_rgs  
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/raw/a549_rgs.bb  
shortLabel a549_rgs  
longLabel A549 raw  
parent universalepi_hic off  
subGroups dataType=raw cellType=A549  
visibility hide

track a549_asc  
bigDataUrl https://boevalab.inf.ethz.ch/resources/universalepi_pred_encode/raw/a549_asc.bb  
shortLabel a549_asc  
longLabel A549 raw  
parent universalepi_hic off  
subGroups dataType=raw cellType=A549  
visibility hide

You can add more detail to the longLabel so that the two tracks are easier to distinguish, for example:

longLabel A549 rgs raw
longLabel A549 asc raw

We have added the hub to our list of Public Hubs: https://genome.ucsc.edu/cgi-bin/hgHubConnect?#publicHubs

We traditionally announce new public hubs. Below is a draft announcement, along with a condensed version for Bluesky/X. Let us know if any information seems inaccurate, if you have any tags you would like us to include, or if the credits statement properly acknowledges the relevant group(s).

Full announcement draft:

We are pleased to announce the release of a new Public Hub, UniversalEPI ENCODE for hg38. This track hub displays normalized ATAC-seq and Hi-C interaction predictions based on ENCODE data and generated using UniversalEPI. UniversalEPI is an attention-based deep ensemble model that predicts enhancer–promoter interactions (EPIs) up to 2 Mb apart using only DNA sequence and chromatin accessibility (ATAC-seq) data. The hub contains two composite tracks:

Hi-C Predictions - This composite track displays Hi-C interaction predictions generated by UniversalEPI using normalized ATAC-seq data from ENCODE cell lines and primary cells as input. Two tracks are provided per input ATAC-seq dataset: one log-transformed ICE-normalized Hi-C track and one z-score Hi-C track.

Normalized ATAC-seq - This composite track shows the normalized ATAC-seq data used as input for UniversalEPI to predict Hi-C interactions. The data is derived from ENCODE cell lines and primary cells, offering a broad view of chromatin accessibility across various cell types.

We would like to thank the Boeva Lab at ETH Zurich for creating this hub.

Social media condensed version draft :

We are pleased to announce our latest public hub, UniversalEPI ENCODE for hg38. It shows ATAC-seq and Hi-C interaction predictions based on ENCODE data, generated by UniversalEPI.

Thanks to the Boeva Lab at ETH Zurich for creating this hub.

I hope this is helpful. Please include gen...@soe.ucsc.edu in any replies to ensure visibility by the team. All messages sent to that address are archived on a publicly-accessible Google Groups forum. If your question includes sensitive data, you may send it instead to genom...@soe.ucsc.edu.

Gerardo Perez
UCSC Genomics Institute

Grover Aayush

unread,

Jul 31, 2025, 11:56:08 AMJul 31

to Gerardo Perez, Valentina Boeva, gen...@soe.ucsc.edu

Dear Gerardo,

Thank you for the suggestions. Glad to see that the UniversalEPI predictions are now available as a public hub.

I agree that the current visualization looks very dense. Is it possible to configure the current bigInteract files in trackdb such that they show interactions with “both ends in window” by default? In this case, it looks somewhat better (see below). Otherwise, is it possible to apply a stronger filter by default in trackdb without changing the input file i.e. we can show all interactions greater than score of 200 by default but the user can choose to set the threshold to 0 in the track settings if needed? If these cannot be applied to bigInteract, we will be happy to consider converting the files to hic.

We have also updated the longLabel for different experiments of the same cell lines, as suggested.

Finally, announcement drafts read well! In the condensed version, it might seem like UniversalEPI predicts both Hi-C and ATAC-seq, which might be misleading so I would suggest the following for the social media condensed version:

“We are pleased to announce our latest public hub, UniversalEPI ENCODE for hg38. It shows Hi-C interaction predictions based on ENCODE ATAC-seq data, generated by UniversalEPI.

Thanks to the Boeva Lab at ETH Zurich for creating this hub.”

Everything else looks great to me! Once again, thank you for the support in making our predictions publicly available.

Many thanks,

Aayush

From: Gerardo Perez <gpe...@ucsc.edu>
Date: Tuesday, 29 July 2025 at 18:15
To: Grover Aayush <agr...@ethz.ch>
Cc: Valentina Boeva <valenti...@inf.ethz.ch>, "gen...@soe.ucsc.edu" <gen...@soe.ucsc.edu>
Subject: Re: [genome] Re: Track hubs

Hello, Aayush.

Thanks for the explanation. The current display feels very dense, with too many connections. It might help to apply stronger filters, use the hic format, or try a filtered version with hic. Let us know if you would like to explore that option.

We did notice that some tracks share the same longLabel, which can be confusing when displayed in the browser graphic.

Gerardo Perez

unread,

Aug 5, 2025, 4:53:05 PMAug 5

to Grover Aayush, Valentina Boeva, gen...@soe.ucsc.edu

Hello, Aayush.

Yes, you can add the endsVisible two setting to the trackDb.txt file to enable the “both ends in window” display by default. To apply a stronger filter, you can use the scoreFilter setting, for example, adding scoreFilter 200 will display only interactions with a score greater than 200 by default. Users can still adjust this threshold in the track settings if needed.

Thank you also for the feedback on the announcement. We will go ahead and use your suggested condensed version for the social media post:

We are pleased to announce our latest public hub, UniversalEPI ENCODE for hg38. It shows Hi-C interaction predictions based on ENCODE ATAC-seq data, generated by UniversalEPI: link to hub

Thanks to the Boeva Lab at ETH Zurich for creating this hub.

We are planning to send out the release announcement tomorrow (08/05/2025).

We appreciate the refinements you have made to the hub and your efforts to make it publicly available.

I hope this is helpful. Please include gen...@soe.ucsc.edu in any replies to ensure visibility by the team. All messages sent to that address are archived on a publicly-accessible Google Groups forum. If your question includes sensitive data, you may send it instead to genom...@soe.ucsc.edu.

Gerardo Perez
UCSC Genomics Institute

Grover Aayush

unread,

Aug 5, 2025, 7:55:52 PMAug 5

to Gerardo Perez, Valentina Boeva, gen...@soe.ucsc.edu

Hi Gerardo,

Thank you for the suggestions. I have now modified the trackDb.txt with “endsVisible two” and “scoreFilter 200”.

Looking forward to announcement tomorrow.

In the trackDb file, you could add subGroup2 and dimensions settings to the universalepi_hic composite track, such as the following:

Grover Aayush

unread,

Aug 6, 2025, 12:36:00 PMAug 6

to Gerardo Perez, Valentina Boeva, gen...@soe.ucsc.edu

Hi Gerardo,

Thank you for the suggestions. I have now modified the trackDb.txt with “endsVisible two” and “scoreFilter 200”.

Looking forward to announcement tomorrow.

Many thanks,

Aayush

From: Gerardo Perez <gpe...@ucsc.edu>

Date: Tuesday, 5 August 2025 at 22:53

In the trackDb file, you could add subGroup2 and dimensions settings to the universalepi_hic composite track, such as the following:

Reply all

Reply to author

Forward