Public Hub QA

From Genecats
Jump to navigationJump to search

Overview

Public hubs are track or assembly hubs contributed by the worldwide research community. Public Hubs are wrangled and QA'd by the UCSC GB QA team. The QAer should work directly with the data submitter from the start to get the track hub in the correct format, etc. Then the QA person QA's the hub (a light QA compared with native tracks) and releases it. QA communicates directly with the data contributor throughout the process. If QAer needs technical help, please contact the Project Manager or an engineer.

Public hubs are made visible by a line in the hubPublic table that the QA-er will add to the various hgcentral* databases. For example:

+----------------------------------------------+-------------------+---------------------------------------+---------------------+---------+--------+--------------------------------+
| hubUrl                                       | shortLabel        | longLabel                             | registrationTime    | dbCount | dbList | descriptionUrl                 |
+----------------------------------------------+-------------------+---------------------------------------+---------------------+---------+--------+--------------------------------+
| http://lisanwanglab.org/DASHR/tracks/hub.txt | DASHR small ncRNA | DASHR Human non-coding RNA annotation | 2015-12-20 11:30:47 |       1 | hg19,  | http://lisanwanglab.org/DASHR/ |
+----------------------------------------------+-------------------+---------------------------------------+---------------------+---------+--------+--------------------------------+

Automation Script

After you do manual QA of the hub tracks, description pages, and the hub passes the requirements, you can run all the table insertion steps below with the following script. Hit "N" for the first question and read carefully. Note that this script can insert a new hub OR update an old one. Follow the prompts.

~/genecats/qa/testTools/hubPublicScripts/updateHubPublic

Contact Daniel Schmelter or edit the script yourself for bugs/improvements.

Public Hub QA Process

The QA process for public hubs is simple: (1) Check that the hub meets our required guidelines, (2) if so, add it to the public list; if not, work with the contributor to ensure it meets those guidelines. Each section below contains more details as to what steps of the QA process you carry out on each of our sites/machines. If applicable, recommend that the hub creator review the metadata guide to add any extra information for their experiments.

On hgwdev

This is where most of the Public Hub QA process will happen.

1) Add Hub to the hubPublic table on hgwdev

First, run hubPublicCheck on your hub:

 hubPublicCheck hubPublic -udcDir=. -addHub=http://lisanwanglab.org/DASHR/tracks/hub.txt 

The option -udcDir= is need to prevent people from using the default udcCache directory -- we all need to use separate -udcDir in order to avoid stepping on each others' toes.

This command will generate the MySQL insert statement needed to add this hub to the hubPublic table. Here is an example command output by hubPublicCheck:

insert into hubPublic (hubUrl,shortLabel,longLabel,registrationTime,dbCount,dbList) values ("http://zlab.umassmed.edu/zlab/publications/UMassMedZHub/hub.txt","UMassMed ZHub", "UMassMed H3K4me3 ChIP-seq data for Autistic brains",now(),1, "hg19,");

Then, use hgsql to insert the line for this hub into the hubPublic table on dev (or use Daniel's above script to update dev/beta/RR):

hgsql -e '<your hubPublicCheck command here>' hgcentraltest

2) Display the hub in the Genome Browser and do some minimal QA

After adding the new hub to the hubPublic table, connect the hub and view in the Genome Browser. The primary QA you will do is ensuring that the hub meets our required guidelines, such as checking for track descriptions with contact information.

Based on the data, you should explore whether a data set should overlap or avoid certain regions such as coding exons (or maybe it would be promoters, or areas of open chromatin, or common SNPs, or highly conserved regions, or 5’ UTRs, or 3’ UTRs, or mitochondria, or the sex chromosomes, depending on the type of data, you can ask the hub provider for input to understand expectations of the data if it isn't clear). Using an idea of what the data is describing, try to identify at least one native track in the Genome Browser that one might expect to correlate or anti-correlate with the tracks in the hub dataset and visually spot check a few.

You should also look for recommended guidelines that the hub violates. For these recommended guidelines, note those that if fixed would greatly improve the usability of the hub for our users. For example, if a hub contains 300 tracks, but they aren’t organized into composites or superTracks, you should recommend that they group their tracks in a reasonable manner.

Notes for Assembly Hubs

In addition to the normal public track hub QA, there are a few things you should pay attention to:

The QA for a public Assembly Hub is very similar to that of a public Track Hub, although there are a few things unique to an Assembly Hub.

  • Check that the 2bit files exist and that the genomes.txt file points to the correct 2bit
  • Make sure the correct labels show up in drop down menus on the gateway page
  • Make sure contact information is clearly displayed on all description pages (gateway description and track description pages)
  • Warning: Be sure to check all your assemblies on the RR before you release. We have a lot of unreleased assemblies on hgwdev, a hub developer once developed a hub with preview assemblies, which worked fine on hgwdev, but on the RR such assemblies failed as the data wasn't there.
    • Also note that some hub developers might try to sneak in an assembly hub without understanding them correctly, see #20761 where genome hub_10649_araTha1 was used to try to reference another assembly hub (someday we might support this).

3) Pass feedback to the hub contributor

If during the previous step, you encountered issues with the hub or noticed that it violates some of our required guidelines, pass these on to the hub contributor in an email. When contacting the hub contributor, lay out your feedback in a clear and concise manner, such as through a numbered list. Often it’s helpful to not just point out the issues but to provide a solution as well, especially if it’s the misuse of a trackDb tag.

For our “recommended” guidelines, you should only pass on feedback for those items that would greatly increase the usability of the hub. For example, organizing hundreds of loose tracks into a superTrack or composite. When passing along these you should note that the contributor isn’t required to change these things, but that it would greatly increase the usefulness of their hub.

On hgwbeta

1) Add the hub to the hubPublic table on hgwbeta

Same as the step for adding this hub to hubPublic on hgwdev. Note the quotation marks (' ' vs " "), you will want to use single quotes(' ') because the output of hubPublicCheck is encapsulated in double quotes (" ")

hgsql -h hgwbeta -e '<insert hubPublicCheck output>' hgcentralbeta

2) Cursory QA on beta

Be sure that the hub and all its tracks load properly on beta. The best way to check this is by clicking the “hide all” button on hgTracks and then navigating to the “Configure” page. On the Configure page, click the “show all” button on the track group for your track hub and then click “submit”. Check that all of the tracks load and that you don’t see any yellow error messages indicating that there were issues loading certain tracks.

Release to the RR

Use the same insert statement that you used to add this hub to hubPublic on hgwbeta to add this hub to the hubPublic table on the RR.

hgsql -h genome-centdb -e '<insert hubPublicCheck output>' hgcentral

Note: If your hub has restricted data (data only loading on the IPs of certain machines) be sure the Public Hub Provider is given all the IPs of our mirrors:

128.114.119.* = genome.ucsc.edu
129.70.40.99 = european mirror, genome-euro.ucsc.edu
134.160.84.67 = asian mirror, genome-asia.ucsc.edu
 128.114.198.32 = genome-test.gi.ucsc.edu, used by developers and for debugging

Post-RR release

1) Rebuild and push hub search files and tables (Depreciated mid-2021)

These steps are now done automatically over each weekend and can be ignored by QA. They serve as a reference for what once was a time-consuming, manual process.

With new Public Hubs (especially with descriptionUrls), once they are on the beta/RR, be sure to build and push an update of the index files.
A. To build these files navigate to hive and run the doPublicCrawl script (you might want to use nohup to let this run in the background). The script took approximately 9 hours on 3/10/21.

cd /hive/groups/browser/hubCrawl
nohup ./doPublicCrawl  

B. To make life easier for the admins also request a separate email to push a table:

SUBJECT:Push hub search table to hgwbeta
Please push the following table 

hubSearchText

from the hgcentraltest database on hgwdev to the hgcentralbeta database on hgwbeta

After the table has been pushed, please 'flush tables' on hgwbeta. 

Reason: Updating the public hub search table on hgwbeta, refs #

C. On Beta you should now be able to search parts of the text on the new hub's descriptionUrl, hub's short or long label, assembly, or track labels. For example a search of methpipe matches a line on the DNA Methylation descriptionUrl: http://smithlabresearch.org/software/methbase/ or hg38 pulls up all the hg38 hubs.

D. To make life easier for the admins also request a separate email to push a table:

SUBJECT:  Push hubSearchText table to RR
Hello pushers,

Please push the following table:

hubSearchText

in the hgcentralbeta database on hgwbeta to hgcentral on genome-centdb/euro/asia.
After the table has been pushed, please 'flush tables' on genome-centdb/euro/asia.

Reason: Updating the hub search table on the RR, refs #

2) Notify hub contributor

Contact the hub contributor and let them know that they can contact our internal mailing list (genome-www) with any questions or concerns.

3) Create a Public Session and Image caption for announcements

Create a snapshot image that you'll use on a Twitter/Facebook post and also create a Public Session with a very short description.

Example post for a hub on Twitter:

Thanks to @PeteHaitch, @LindsayRizzardi, @KasperDHansen and others at Johns Hopkins for the Brain Epigenome public hub, now available!
http://genome.ucsc.edu/cgi-bin/hgTracks?hgS_doOtherUser=submit&hgS_otherUserName=chmalee&hgS_otherUserSessionName=hg19_brainEpigenome

Example post for a hub Public Session:

Description: This session highlights data from the Brain Epigenome Hub, created by Peter Hickey, Lindsay Rizzardi, and Kaspar Hansen and others at Johns Hopkins University. The hub shows methylation, ATAC-seq, and RNAseq across different brain regions.

4) Send genome-announce email

Here are some previous examples of announcement emails for public hubs. It is an opportunity to share a sentence or two about the lab and data (and maybe thank them for creating the public hub). The news could even be tweeted and added to Facebook too.

5) Add your hubUrl and email to automatic uptime emailer

Log into qateam@hgwdev and add your hub to the bottom of the auto mailer cron file:

/cluster/home/qateam/cronScripts/hubPublicMailStatus.tab

This system is an automatic hubUrl checker that emails the contributor's email after 2 days of being down, no more than once every 6 days. It is run out of qateam's hgwdev crontab.

Uncommon situation: Changing the URL in an Existing Public Hub

If a hub provider asks to change their Public hub's URL to a new address, it breaks any saved sessions and cart connections to that Public Hub. This is because the hubStatus ID is unique to each URL and needs to be updated in the SQL table. Needs work here to fix the hubStatus table.

Fortunately, the actual hubPublic url change [not the hubStatus change] has been scripted and can be found in the following directory:

/cluster/home/dschmelt/danScripts/hubPublicUrlChanger

The script requires no arguments and has info requests and asks for approval before making any changes.

Note: edits to hubStatus can have drastic consequences upon the RR, before making any edits be sure to check with senior team members and also create a backup of hubStatus by downloading the table with a SQL select command (if in doubt, do not edit hubStatus).

QA for UCSC-hosted public hubs

Public hubs created by browser engineers are an alternative to native tracks for specialty data. These types of hubs are QAed and released in a very similar manner to externally hosted hubs with a few minor differences. The main difference is that there is an additional step to push the hub files from the development system to the public download server. UCSC-hosted public hubs were previously hosted on hgwdev in /gbdb/hubs and then pushed to hgdownload for display on the public site. Now we host these hubs on hgwdev in /usr/local/apache/htdocs-hgdownload/hubs/ although they are still pushed to hgdownload for display on the public site. This shift was done to reduce confusion with other /gbdb pushed that normally go to hgnfs1. In an email thread with cluster-admin, Hiram wrote:

Perhaps the confusion arises because we are using the directory /gbdb/hubs/
on hgwdev to construct our symlinks to release these files.

These items have nothing to do with /gbdb/, the hubs are not
under /gbdb/ on hgdownload, they are /hubs/ 

On hgwdev they are under /gbdb/hubs/ only because /gbdb/ is the location we use to
create symlinks to deliver stuff to the outside world. 

We could instead place the symlinks on hgwdev in the directory:
  /usr/local/apache/htdocs-hgdownload/hubs/

And thus eliminate any reference to gbdb

You should still review the hub on hgwdev as described above. Since we’re the ones hosting and providing these hubs, it’s alright to be a little more strict in regards to our hub guidelines. Once the hub is looking good on hgwdev, you can release it to the RR using the steps described in the next section.

Releasing UCSC-Hosted big data Public Hubs

This step is rare and only relevant for a big data hubs that UCSC is hosting.

These hubs are made available with a push request from /usr/local/apache/htdocs-hgdownload/hubs/ on hgwdev to /mirrordata/hubs on hgdownload.

Here's an example push request:

Please push the following file:

/usr/local/apache/htdocs-hgdownload/hubs/newHub/*

from hgwdev --> hgdownload/hgdownload-sd
    (in path, "/usr/local/apache/htdocs-hgdownload/" should become "/mirrordata/" on hgdownload)

Note that items that are symlinked on hgwdev should become real files on hgdownload. 

Reason:  Releasing new UCSC hosted hub newHub to hgdownload.

Thanks!

It may be useful to note that the UCSC GTEx data have restrictions on access. Apache for hgdownload only allows RR to access the hub data files so the hub displays on the RR only (the hub won't load on other sites, and files can not be directly downloaded -other external Public Hubs have taken similar steps to control their data).

What to do if a Public Hub is down?

If you notice that a hub is consistently down for an extended period of time (3-5 days), then you should contact the hub contributor to let them know that their hub is having issues. We keep a page of the contact information for all of our public hubs here: http://genecats.soe.ucsc.edu/qa/test-results/publicHubContactInfo/publicHubContact.html. We also have a cronjob that checks the status of all of our public hubs, so be sure to check in with the person receiving those emails before sending your message.

We now have a continuously updating log of down public hubs. It's on the qateam dev login. Check it out here:

[qateam@hgwdev /cluster/home/qateam] vi /hive/users/qateam/perf/hubPublicCheckCron.log


Hub Public Coordinator Role

One QAer historically has been assigned the role of Hub Public Coordinator. This role is to maintain an updated and error-free listing of Public Hubs by informing hub providers when their hub hosting sites go down. This role has changed a lot from 2018 to 2022 during Dan's tenure. It is now almost completely automated and on the qateam cron at "~/genecats/qa/crontabs/hgwdev.crontab". I will describe the two-part automation below.

1. First is keeping hubs updated: labels/titles and the search index. These used to be manual tasks and are now fully automated.

  • The hubPublic labels/titles are updated by the script "/cluster/bin/scripts/hubPublicAutoUpdate", which runs on each of our 3 sites every weekday morning from the QAteam crontab.
  • The hubSearchText search index is updated on Dev by the script "/cluster/bin/scripts/hubSearchUpdate" every Tue/Sat and is auto-pushed by the Pushers every Sunday.

2. Second is keeping the hubs error-free. If a hub.txt website link is no longer accessible, you'll see an error in the Public Hub listing in red. Historically, these errors were emailed to a certain person and they'd wait about a week before contacting the hub provider.

Now, this is automated and hubs are checked every 2 hours and if they get errors for 24 consecutive checks (48hrs) then they get an automatic email. Here is the script that runs that auto-email: "/cluster/bin/x86_64/hubPublicMail". This is also run out of the QA Team crontab. This program will spam the hubContact email every few days until they fix it. If a hub provider doesn't fix their error-prone hub after perhaps 1 month, the hub should be removed from the hubPublic listing.

For reference, it is worth noting that we have not historically heckled hub providers over broken tracks (which there are many), only if their overall hub is down. Also, the hub contact list is here:

https://genecats.gi.ucsc.edu/qa/test-results/publicHubContactInfo/publicHubContact.html

As the person in charge, you should be a MAILTO on each of these 3 cronjobs. Best of luck!