New track checklist

This is the checklist for QA to follow when releasing Genome Browser tracks. General rule: stop testing when you come to the first non-trivial error, and bounce the track to the B queue until the track sponsor fixes it.

Claiming a track: pushQ and redmine

Once you've decided to QA a track you will need to put your name down as the "reviewer" in the related pushQ entry:

http://hgwbeta.cse.ucsc.edu/cgi-bin/qaPushQ

Note that you will need to check and/or "mark as checked" the following pushQ items:

• sizes: click the "show sizes" button to check that the files and tables associated with the track have sizes that make sense
• makedoc: mark this as "yes" once you've checked the makedoc (see below)
• joiner: mark this as "yes" once you've run joinerCheck (see below) or "X" if this track is ignored
• index: we no longer check this

Assign the redmine ticket to yourself and set the status to "reviewing". For more info on the redmine workflow see this wiki: http://redmine.soe.ucsc.edu/projects/genomebrowser/wiki/Track

Background familiarity

If you haven't worked with a track like this one, and it already exists on the RR for another assembly (use 'getAssemblies.csh tablename' on hgwdev to check), use it a bit on the RR to become familar with it. For complex tracks, look for genomewiki pages on it, and/or look through the push queue for previous problems encountered on similar tracks. Familiarize yourself with the tables and files needed for this track.

Makedoc

Look for an entry for your track in src/hg/makeDb/doc/$db.txt. If there is no mention of this track, make a request to the track sponsor. Basic tracks that are loaded automatically with a new assembly may only have a tiny reference in the makedoc. UCSC Genes has its own document.ls

All Ensembl tables are generated by the Ensembl update procedure which is maintained in doc/makeEnsembl.txt without mention of explicit tracks or tables.

In the makeDb/doc/ directory there are some ENCODE specific folders and .txt files to look at, for example encodeAwgHg19.txt that has the Analysis Working Group based uniform processing makedoc information.

• The following tables don't need to be referenced in the makedoc: chromInfo, gap, gc5BaseBw, gold, grp, hgFindSpec, history, trackDb

Release Log and URL

The "Release Log" field in the push queue should generally contain the shortLabel of the track, or a short description of what is being released if it does not make sense to put a shortLabel here. The entry in this field will be added to the auto-generated release log page. If it makes sense to link to the hgTrackUi page for this track, make an entry in the "Release Log URL" field like so:

 ../../cgi-bin/hgTrackUi?db=[DATABASE]&g=[TRACK]

An actual example:

 ../../cgi-bin/hgTrackUi?db=hg19&g=affyU133

An entry is not always required, for example, for tracks in a new assembly.

The BIG QA Script: qaGbTracks

You can use qaGbTracks to run many of the qa scripts in one go on hgwdev, including:

 checks for underscores in table names
 checks for the existence of table descriptions
 checks shortLabel and longLabel length
 positionalTblCheck
 checkTblCoords
 genePredCheck
 pslCheck
 featureBits
 (a version of) countPerChrom

Example: qaGbTracks papHam1 gc5Base output.gc5

Label lengths

Check that the shortLabel is 17 characters or less and that the longLabel is 80 characters or less. The shortLabel is visible in the main hgTracks display if you turn the track to dense. If it is 17 characters or less, it won't be cut off in this part of the display. The longLabel is visible in hgTracks, and it can also be copied from the configuration page (hgTrackUi).

Tables are sorted and internally consistent

Run any of these that apply to your tables:

genePredCheck - Checks that tables in the genePred format are valid. (Tables of this type are usually in the Genes and Gene Prediction group. Usually only the primary table will need this check. Some examples: knownGene, ensGene, refGene.)

pslCheck - Checks that psl tables are valid. (PSL tables show up in nearly any track where alignments are used. Some examples: mrna, est, refSeqAli.)

checkTableCoords - Checks that the genomic coordinates in positional tables are legal (e.g., coordinates are not off the end of a chromosome). If the table passes there will be no output. (Almost every table will need this check, unless it has no chromosome positions in it.)

positionalTblCheck - Checks to see that the a positional table is ordered by chrom and chromStart. A positional table being out of order can cause a huge slowdown in display speed. If the table passes there will be no output. If the table does not pass there will be an error message like:

table hg18.snp129 not sorted starting at row 4867: chr1:387005

Alert the track sponsor if there is an error. He/she may determine that the items are sorted enough to be released. (As long as the items are almost all in order, it will not affect performance.) Also note that Genbank tables are not expected to be in order after updates have run. (Almost every table will need this check.)

FeatureBits and Gaps

Run featureBits, or use the runBits.csh script to run featureBits. runBits.csh checks for coverage and overlap with gap, and also checks for undbridged gaps.

 runBits.csh $db $table

Alert the track sponsor if there are unbridged gaps and this is a track created at UCSC. Put featureBits results in the push queue. If previous assembly also has this track, compare featureBits between current assembly and previous assembly -- if there are big differences between the old and new tracks, alert the track sponsor.

If there is a similar track that this one can be compared to, use either "featureBits -enrichment" or getYield.csh to compare the tracks. Alert the track sponsor if the difference seems unreasonable.

getYield.csh can be used to see how well a new track captures the footprint of existing tracks, such as refGene or xenoRefGene, e.g.,

 getYield.csh hg19 ensGene refGene

output includes:

 yield       = 93.8% (intersection / refGene)
 enrichment  = 24.7x ((intersection / ensGene) / (refGene / genome))

shows that 93.8% of refGene is present in ensGene and that compared to the refGene footprint on the genome, ensGene is 24x enriched for refGenes. (Enrichment is the amount of table1 that covers table2 vs. the amount of table1 that covers the genome. It's how much denser table1 is in table2 than it is genome-wide.)

"featureBits -enrichment" does almost the exact same thing as getYield.csh, except the coverage amount is the number of bases in the intersection of the two tables divided by the first table instead of the second table.

Chromosome coverage

Check the count of items on each chromosome. Bigger chromosomes (the biggest is usually chr1) should have more items. Look for chromosomes that have suspiciously few or no items on them. Note: this script must be run on dev:

 countPerChrom.csh $db $table

or as a histogram:

countPerChrom.csh $db $table histogram
Example: 
countPerChrom.csh hg19 refGene histogram

If a "regular" chrom (not random or haplotype) has no data, a note like this should appear in place of the track drop-down on hgTracks:

 [No data-chr21]

(This is controlled by the "chromosomes" setting in trackDb.ra. The chromosomes line in trackDb.ra specifies the chromosomes that DO have data. This line is the same as the "restrictList" field in the trackDb table.)

OR use these alternative ways to see chromosome coverage:

• Import the table into Genome Graphs.

• Select the table in the Table Browser, hit the "describe table schema" button, and click the "values" link for the chrom field.
  • If the table is very large, there may not be an "info" column to display the "values" link.

joinerCheck

The document kent/src/hg/makeDb/schema/joiner.doc describes what all.joiner is. In a nutshell, all.joiner is a file that describes joinable fields in the UCSC Genome Databases. With each sandbox, alpha, beta build (not make alpha DBS=) all.joiner uses it's definitions of relationships through it's identifiers to link tables. You can see the results in the Table Browser when you click the describe schema button and see the "Connected Tables and Joining Fields" section for tables that have all.joiner definitions. The tool you want to focus on is joinerCheck, which is used to check that the rules in all.joiner are being followed.

Look for your table names in src/hg/makeDb/schema/all.joiner and find the identifiers associated with those tables. Then, for each identifier, run joinerCheck like so:

 joinerCheck -keys -identifier=$identifier -database=$db all.joiner

Note, the above is run from all.jointer's home kent/src/hg/makeDb/schema/, otherwise the location of all.joiner needs to be spelled out in the command line. If there are errors or the table is not mentioned in all.joiner, notify the track sponsor. An entry in the tablesIgnored section of all.joiner is sufficient if there are no table relationships to check.

Be aware of this problem with joinerCheck. Basically, if you get output that looks like this:

 Checking keys on database hg18

that is NOT followed by lines like this:

 anoCar1.blastHg18KG.qName - hits 45332 of 45332 ok

then the rule didn't really run, and you need to remove the -database parameter.

The other option is to run joinerCheck with the verbose option set to 2, for example:

 joinerCheck -keys -identifier=$identifier -database=$db -verbose=2 all.joiner

Unfortunately the the -verbose=2 option outputs all of the identifiers, so you must search through the list to find your identifier. After you find your identifier in the list, you should hopefully see the the output as described above.

You can also run joinercheck with the -times flag:

 joinerCheck -times -database=$db all.joiner

Look for any errors that are relevant to your track.

The runJoiner.csh script is a shortcut for all of the above, but beware that wildcards in tablesIgnored sections are not recognized, and if the problem above occurs, then you need to run joinerCheck directly.

Comparison

Compare your track to a similar track if possible. Look to see that the features in your track are more or less in the same position as similar tracks. Look for a lot of items that have a chromStart or chromEnd that is one position different from existing tracks. This could indicate an off-by-one error in the new track. Also use getYield.csh and/or featureBits to compare tracks (discussed next).

Searching

If the track item names are relatively unique, check to see if search works by pasting an item name in the position/search box. (For example, to check if search is enabled for the "Common SNPs" track, choose a SNP from the track, say "rs17885219," paste it in the box, and hit "jump." If search is enabled, you will either be taken directly to the position of the item, and it will be highlighted in the display, or you will get a list of all of the tracks that contain the item (and your track should be included). One way to find out if your track should be searchable is to use the assembly $db and table name $table in the following command:

hgsql -Ne 'select searchName,searchTable,searchMethod,termRegex from hgFindSpec where searchTable like "%$table%";' $db;

If when searching you get an error, or if you get a list of tracks but your track isn't included in the list, search is not enabled.

If the track item names are relatively similar (e.g., the items in RNA Genes, and TFBS) we don't want to enable search, as it would return too many matches. If search isn't enabled and you think it should be, make that request to the track sponsor.

Finally, if search is enabled, make sure that all of the item names in the track can be found. Do this by checking this page for your assembly and table:

http://genecats.cse.ucsc.edu/qa/test-results/checkHgFindSpec/hgwdevOutput

If your assembly and table appear in this list, there is a problem with searches for some identifiers in the track. You'll see an error message like this one:

 Error: mm9.jaxQtl.name value "Idd21.1" doesn't match termRegex "^[a-z0-9-]+$" for search jaxQtl

This means that the item with the name "Idd21.1" in the jaxQtl (MGI QTL) track on mm9 (mouse) doesn't work. Alert the track sponsor.

Track description

Read the track description and edit for clarity, spelling, and grammar. Be sure our conventions are followed.

Ensure references are in the correct format and in alphabetical order (by first author listed). Links to journal articles should go directly to the journal rather than PubMed if the journal article is open access (i.e., doesn't require a subscription). For articles that are not open access, links can go either to the journal or to PubMed, and they should go to the abstract, not the full text.

Ensure quotes, ampersands, and less than and greater than signs are represented with their html names.

Make sure that any email addresses given on the details page have been through Hiram's sanitizer (encodeEmail.pl). It turns the address into an encrypted HREF "mailto:" address that makes it harder for spammers to use.

All details: 1 data point

Choose a representative data point for the track. Check all details for this data point, including all links. Make sure information from the table is displaying correctly (e.g., if a color is used in the table, make sure that color appears for the item.)

For links that are hard-coded to a particular server, there are some tricks that are used to make them testable on hgwdev and hgwbeta. See the Static_Page_JS_Protocol page for more details.

One method of obtaining an item's data to check is to click the "View table schema" link and use the Sample Rows as entry coordinates to navigate to and review the data.

Another option is to use a MySQL query to pull the information from the table for an item you maybe looking at that you are curious to learn more about. Here is an example query followed by a general method:

hgsql -Ne 'select * from gold where frag like "%AMGL01129756.1%";' oviAri3

hgsql -Ne 'select * from (tableName -displayed in schema) where (fieldName -find appropriate field from schema) like "%(yourSearchTerm -from tableName.fieldName)%";' $db

Performance and Display

For tracks displayed by default, the full chromosome view (chr1) should display within 20 seconds. For tracks which are not displayed by default, the full chromosome view should display within a minute. You can use the 'measureTiming' cart variable to get accurate load times. To enable this timing option, add &measureTiming=0 to the end of your hgTracks URL. If you did it correctly, your URL should look like this: http://hgwdev.soe.ucsc.edu/cgi-bin/hgTracks?db=hg19&position=chr21%3A33296501-33297001&measureTiming=0. The total time it took to load the page will be next to the label 'Overall total time:' somewhere on the page. You can deactivate this timing option by either doing a cartReset or adding 'measureTiming=', with no value specified to your hgTracks URL.

Turn on to full display mode a track that is located physically below your track in the display. Make sure that when your track is in full display mode, that the items in the track below it are still mapping correctly. Sometimes there can be an off-by-one error which is caused by your track. If this is happening, you should not push your track. This is only likely to be an issue with new track types.

Hit the Reverse button and ensure your track displays correctly.

Track Settings (hgTrackUi)

Click on the track name or the mini-button to the left of the track (in hgTracks) to get to the track settings page. Make sure that the track settings work as expected.

Table Descriptions

Hit the "view table schema" button (on hgTrackUi, hgc, or hgTables) and make sure there is a description column present with descriptions of the table fields. If a track has more than one table, be sure to check for table descriptions on each of them. The description column uses the 'tableDescriptions' table to display this information.

The descriptions in the 'tableDescriptions' table are built from autoSql (or .as files), and are added to that table via the script buildTableDescriptions.pl (kent/src/test/buildTableDescriptions.pl). This script first looks for .as files that match a database table by name or trackDb type. If it doesn't find a matching .as that way, it can match .as with table by comparing the set of fields in a .as versus the set of fields in the database table. This means that if one .as has the same fields as several database tables, the .as name doesn't matter -- the script will still match them up. The 'tableDescriptions' table is built nightly on hgwdev and hgwbeta, and must be pushed to the RR if it contains descriptions for a new type of table (things like psl will already be out there). A cron job sends a push request email to the admins to push the tableDescriptions tables to the RR once a week.

Having as few .as files as possible is a good thing because duplicated content is harder to maintain, and hg/lib is already enormous so it's good to reduce the number of new files. To help reduce the number of files added to hg/lib, try to do the following when QA-ing a new track:

• Check tables that are similar to your track's tables to see if there are any where all of the fields are identical to yours
• Check kent/src/hg/lib/ for any .as files that have the same name as yours, or may have been created for your track's tables
• If there are new .as files for your track's tables that are unnecessary, notify the track sponsor that these should be removed

More information on the tableDescriptions table and autoSql can be found here:

• tableDescriptions
• AutoSql

Downloads

If there are files associated with the track that are to be pushed to hgdownload, check to see that there is a README that makes sense and the files have an md5sum.txt file that goes with them and is correct. Check the file itself to make sure it is not corrupted and that it contains what is expected. If it is a gzipped file, you can do "zcat file.gz | head" and "zcat file.gz | tail" to look at it. Looking at the last part of the file can sometimes catch corruption that can't be seen by only looking at the first part.

Data files destined for hgdownload are organized on hgwdev at:

/usr/local/apache/htdocs-hgdownload/goldenPath/*

and can be viewed in a browser from http://hgdownload-test.cse.ucsc.edu/downloads.html. Non-data files (such as downloads.html) are in the "hgdownload" git repository. See the Static_Page_Protocol for instructions on checking out that repository. Push requests for downloads should look like something like this:

Please push files from here on hgwdev:
    /usr/local/apache/htdocs-hgdownload/goldenPath/$db/file
To here on hgdownload:
    /usr/local/apache/htdocs/goldenPath/$db/file
(in the path, "htdocs-hgdownload" should become "htdocs") 

Finally, check to see if the downloads files ought to have a link from downloads.html. If so, add the link and push downloads.html (after the files are already pushed!).

Push to hgwbeta

Note for Ensembl gene tracks:

• Before pushing to beta, use the Ensembl_QA script.

Special case for 'seq' and 'extFile' tables:

• Do not push seq or extFile tables from dev to beta. You must use the copyExtSeqRows.csh script to move only the rows needed.

Notes on existing tracks:

• If this is an update to an to an existing track, you may want to hold off on this step so that you can compare old and new tracks on hgwdev and hgwbeta.
• Open the track on hgwbeta before staging it to make sure that the update won't cause a cart clash for users currently looking at the track (as evidenced by a completely blank screen, for instance). If you need to do a cartReset to get the track to show up correctly, something is wrong.

Push all tables (EXCEPT seq and extFile tables, see note below) from hgwdev to hgwbeta. There are two ways:

 sudo mypush $db $table mysqlbeta

or, for a list of tables:

 bigPush.csh $db $tableListFile

Next, make beta on hgwdev in kent/src/hg/makeDb/trackDb like so:

 make beta DBS=$db

To make beta on more than one db at a time:

 make beta DBS='$db1 $db2 $db3 $db4 etc'

Request a push of any listed supporting files in /gbdb from hgwdev to hgnfs1. Note that hgwbeta and the RR share the files on hgnfs1, so once these files are in place, there is not another push required when the track is released to the RR.

Make Public

Make your track public by using the "make public" command on hgwdev while in the trackDb directory (src/hg/makeDb/trackDb):

   [user@hgwdev trackDb]$ make public DBS=$db

Your track should now be visible on the hgwbeta-public server.

If your track is not visible, you may want to check that your track has the correct release tag. Also see [Three State TrackDb] for more information.

Push Request

When the track is ready to be released, ask admins (push-request at soe) for a push of the tables (including trackDb_public and hgFindSpec_public, if needed) from mysqlbeta to mysqlrr. Push any associated downloads. Push any associated static docs. Notify (cc:) the track sponsor on the push request.

If the track already exists on the RR you may need to selectively push the tables/trackDb. Refer to Replacing old tables with new ones for more information.

If this is a repush of existing data that was found to be problematic, add a note to repush.html. The file to edit is in the genecats tree, at genecats/qa/repush.html. (This doesn't happen very often.)

Validate on the RR

Check your track on the RR. Check that searches work (if not, you probably need to push the hgFindSpec_public table). Also check that all default tracks still display. If you filled in the "Release Log URL" field in the push queue, check the next day to be sure that the link from the release log works as expected.