KnownGene build

Build UniProt and Protein databases

I haven't been doing this recently. We need to look into whether the work Max has done with uniprot should replace this.

JC 8/24/24: I rebuilt UniProt a couple of releases ago - it took a very long time because it's much larger now. I didn't get around to rebuilding Protein. Replacing this with content from the uniprot otto job is a good idea, but there may be some additional tables that we'll still need to construct from the otto data. I know accToTaxon is used by hgLinkIn (used by UniProt to generate links to our site), and that's not part of the otto build.

Consider updating underlying databases

• If it's been a while since we updated BioCyc versions for this species, consider doing so (we download the public BioCyc flat file .tar.gz distribution for mouse and human)

JC 8/14/24: Last I checked BioCyc, they were no longer providing any public data files for mouse or human. The only data they provide for free is EcoCyc (e. coli). We can try reaching out to them about a free license for our display and link purposes, but their site seems pretty explicit that data access requires a paid subscription.

Initialize work directory

• Set version variable

 export GENCODE_VERSION=V39

• Start a screen.

 screen -S knownGene$GENCODE_VERSION

• Create and cd into work directory of the form /hive/data/genomes/$db/bed/gencode$GENCODE_VERSION/build

   export db=hg38
   mkdir /hive/data/genomes/$db/bed/gencode$GENCODE_VERSION/build
   cd /hive/data/genomes/$db/bed/gencode$GENCODE_VERSION/build

• Set PATH to include $HOME/kent/src/hg/utils/otto/knownGene

 PATH=$HOME/kent/src/hg/utils/otto/knownGene":$PATH"

• Copy buildEnv.sh from previous build on this db

  olddir=`ls -trd /hive/data/genomes/$db/bed/gencodeVM*/build | tail -n 2 | head -1`
  cp $olddir/buildEnv.sh  buildEnv.sh
  edit buildEnv.sh to have correct values
  . buildEnv.sh

• Find Table and File list from previous build

  cp ${oldGeneDir}/${PREV_GENCODE_VERSION}.files.txt  .
  cp ${oldGeneDir}/${PREV_GENCODE_VERSION}.tables.txt  .

• Confirm existing assembly tables are in a knownGene* database (sort syntax is a bashism - if using tcsh, sort the tables before the diff)

  hgsql ${oldKnownDb} -Ne "show tables" > ${oldKnownDb}.tables.txt
  diff <(sort ${PREV_GENCODE_VERSION}.tables.txt) <(sort ${oldKnownDb}.tables.txt)

Setting environment variables

The environment variables used in the build are set in the script buildEnv.sh. All the other scripts assume that this script has been sourced in the current shell. You have to edit this by hand. Most of the variables don't change. The hairiest ones are the other assemblies for the blast tables.

Running the build

To run the build execute hg/utils/otto/knownGene/buildKnown.sh.

  buildKnown.sh &
  tail -f doKnown.log

It builds into the knownGene${GENCODE_VERSION} database. It does the following steps:

• Extracting Gencode data
• Building initial knownGene table
• Adding primary reference tables
• Building final knownGene core tables
• Building bigGenePred
• Building GTF file

NB: The GTF step is not currently a part of the script. It can be manually run with

 cd /hive/data/genomes/$db/goldenPath/bigZips/genes
 genePredToGtf -utr ${tempDb} knownGene ${db}.knownGene.gtf
 rm -f ${db}.knownGene.gtf.gz
 gzip ${db}.knownGene.gtf

Copying over tables

drop chromInfo and history from knownGene database

  hgsql knownGene${GENCODE_VERSION} -Ne "drop table if exists chromInfo, history"
  hgsql knownGene${GENCODE_VERSION} -Ne "show tables" | egrep "knownGene|kgXref" > ${GENCODE_VERSION}.tables.txt
  hgsql knownGene${GENCODE_VERSION} -Ne "show tables" | egrep -v  "knownGene|kgXref" >> ${GENCODE_VERSION}.tables.txt

look for unexpected differences between this release and the last one

  diff ${PREV_GENCODE_VERSION}.tables.txt ${GENCODE_VERSION}.tables.txt

drop old tables

  hgsql $db -Ne "drop table knownGene, kgXref;"
  grep -v "ToKg" ${PREV_GENCODE_VERSION}.tables.txt | egrep -vw "knownGene|kgXref"  | awk '{printf "drop table %s;\n", $1}' > toDrop.lst
  cat toDrop.lst | hgsql $db

check for orphans and drop them (or build them) if appropriate

 hgsql $db -Ne "show tables like 'known%'"  > orphan.lst

copy tables from knownGene database to assembly database

 copyFilesToAssembly.sh ${GENCODE_VERSION}.tables.txt knownGene${GENCODE_VERSION} > copyScript.txt
 cat copyScript.txt | hgsql $db

Edit trackDb to add new trackDb

 cd $HOME/kent/src/hg/makeDb/trackDb/*/$db
 vi trackDb.ra
     include knownGene.ra beta,public
     include knownGene.alpha.ra alpha
 sed "s/$PREV_GENCODE_VERSION/$GENCODE_VERSION/g" knownGene.ra > knownGene.alpha.ra
 cp knownGene$PREV_GENCODE_VERSION.html knownGene$GENCODE_VERSION.html
 git add knownGene.alpha.ra knownGene$GENCODE_VERSION.html trackDb.ra
 git commit -m "$GENCODE_VERSION knownGene trackDb"
 git push
 cd ../..
 make DBS=$db alpha
 cd $dir

NB: In the above process, don't forget to edit the new knownGene*.html file before committing it. All mentions of the version number will need to change, the statistics will need to be updated with the ones posted for this release on the GENCODE website ("Immunoglobulin/T-cell receptor gene segments" is the sum of the two listed values), and the references section might need to be updated if there's a new paper that should be cited. Don't forget to amend the credits section too, as needed.

Then edit knownGeneArchive.ra, copying in the settings from the previous version of knownGene as a new subtrack. Be sure to adjust settings as appropriate for the release (parent, priority, externalDb). Remember to also set it to release alpha (the same release status as the new knownGene track) unless the entire archive track set is set to release alpha (it might be included that way in trackDb.ra).

Adding IsPcr server

On hgwdev, drop old records in blatServers and targetDb.

 hgsql hgcentraltest -Ne "delete from blatServers where db like '${db}Kg%'"
 hgsql hgcentraltest -Ne "delete from targetDb where name like '${db}Kg%'"

Ask cluster-admin to start an untranslated, -stepSize=5 gfServer on /gbdb/$db/targetDb/${db}KgSeq${GENCODE_VERSION}.2bit

  genIspcrMail.sh

send to cluster-admin

cluster-admin will say something like this:

 Starting untrans gfServer for mm39KgSeqV38 on host blat1b port 17921

where blat1b is the serverName and the port is 17921

Add this info to blatServers and targetDb tables in hgcentral.

addIspcrToCentral.sh serverName port

all.joiner changes

I haven't added anything to this recently.

The relevant id's are :

knownGeneId

  joinerCheck all.joiner -identifier=knownGeneId -keys  -database=${db}

Bundle up logs and check them in

Make a short log file in makeDb/doc/ucscGenes/.

Redmine ticket files and tables

NB: You'll probably want to generate the ${GENCODE_VERSION}.files.txt file, since there's nothing in the above procedure that does it. Unless the list of files changes, the easiest way is to just make a copy of the previous version's file and update the version numbers inside it (it's like 6 or 7 lines).

Post release push "other species" blast tables

Load the other species blastTab tables.

   buildLoadOther.sh