17748af09f0add98009e2c4c181cb4e75d068d26 hiram Wed Sep 16 14:17:15 2020 -0700 modified procedure to make a new style LRG transcript track refs #24672 diff --git src/hg/makeDb/doc/hg38/hg38.txt src/hg/makeDb/doc/hg38/hg38.txt index 88a1e90..6b29194 100644 --- src/hg/makeDb/doc/hg38/hg38.txt +++ src/hg/makeDb/doc/hg38/hg38.txt @@ -3643,66 +3643,94 @@ -dbHost=hgwdev -workhorse=hgwdev monDom5 hg38) > rbest.log 2>&1 # real 90m56.189s _EOF_ ############################################################################# # LOCUS REFERENCE GENOMIC (LRG) REGIONS AND TRANSCRIPTS (DONE 10/25/19 angie) # Redmine #13359, #24285 -- otto-mate To Do #17877 # previously done 7/7/14, 9/9/16, 5/30/18 set today = `date +%Y_%m_%d` mkdir -p /hive/data/genomes/hg38/bed/lrg/$today cd /hive/data/genomes/hg38/bed/lrg/$today wget ftp://ftp.ebi.ac.uk/pub/databases/lrgex/LRG_public_xml_files.zip unzip LRG_public_xml_files.zip # Run script to convert LRG*.xml files to BED+ for regions and genePredExt+fa for transcripts: + # parseLrgXml.pl updated 2020-09-16 to add four new fields to the gp output + # the four extra fields are identifiers for: + # NCBI transcript, Ensembl transcript, NCBI protein, Ensembl protein + ~/kent/src/hg/utils/automation/parseLrgXml.pl GRCh38 genePredCheck lrgTranscriptsUnmapped.gp #Error: lrgTranscriptsUnmapped.gp:765: LRG_7t1 no exonFrame on CDS exon 46 #checked: 1029 failed: 1 # If there are complaints e.g. about exonFrame, look for inconsistencies in the # affected transcript's coding_region/coordinates vs. exon/intron info in xml. # Contact Variation team leader Fiona Cunningham @EBI to resolve in the background # (missing exonFrame info doesn't affect our track representation because we end up using # psl). We agreed to disagree about exon 46 of LRG_7t1 because that last coding exon # portion is only the stop codon. # No longer necessary to filter out alt and fix patches since they have been added to hg38. + # four extra columns have been added to the genePred (2020-09-16 - Hiram) + # extract them so they can be added to the psl: + awk -F$'\t' '{printf "%s\t%s\t%s\t%s\t%s\t%s %s %s %s\n", $1,$16,$17,$18,$19, $16,$18,$17,$19}' lrgTranscriptsUnmapped.gp | sort > lrgTransExtraFields.tsv + + # the four extra fields are identifiers for: + # NCBI transcript, Ensembl transcript, NCBI protein, Ensembl protein + # Load LRG regions: bedToBigBed lrg.bed /hive/data/genomes/hg38/chrom.sizes lrg.bb \ -tab -type=bed12+ -as=$HOME/kent/src/hg/lib/lrg.as -extraIndex=name ln -sf `pwd`/lrg.bb /gbdb/hg38/bbi/lrg.bb hgBbiDbLink hg38 lrg /gbdb/hg38/bbi/lrg.bb # Map LRG fixed_annotation transcripts from LRG coords to hg38 coords (HT MarkD): lrgToPsl lrg.bed /hive/data/genomes/hg38/chrom.sizes lrg.psl pslCheck lrg.psl #checked: 919 failed: 0 errors: 0 awk '{print $10 "\t" $11;}' lrg.psl > lrg.sizes genePredToFakePsl -chromSize=lrg.sizes placeholder \ lrgTranscriptsUnmapped.gp lrgTranscriptsFakePsl.psl lrgTranscripts.cds pslMap lrgTranscriptsFakePsl.psl lrg.psl lrgTranscriptsHg38.psl mrnaToGene -genePredExt -cdsFile=lrgTranscripts.cds -keepInvalid \ lrgTranscriptsHg38.psl lrgTranscriptsHg38NoName2.gp #Warning: no CDS for LRG_163t1 #Warning: no CDS for LRG_347t1 # It's OK if mrnaToGene complains about "no CDS" for a non-coding tx (RefSeq accession NR_*). grep -l NR_ LRG_163.xml LRG_347.xml #LRG_163.xml #LRG_347.xml + # construct bigPsl with four extra fields + pslToBigPsl lrgTranscriptsHg38.psl bigPsl.txt + + # add the four extra identifiers to the bigPsl file: + join -t$'\t' -1 4 \ + -o 1.1,1.2,1.3,1.4,1.5,1.6,1.7,1.8,1.9,1.10,1.11,1.12,1.13,1.14,1.15\ +,1.16,1.17,1.18,1.19,1.20,1.21,1.22,1.23,1.24,1.25,2.2,2.3,2.4,2.5,2.6 \ + <(sort -k4 bigPsl.txt) lrgTransExtraFields.tsv \ + | sort -k1,1 -k2,2n > lrgExtraTranscriptsHg38.bigPsl.bed + + bedToBigBed -as=bigPsl+5.as -type=bed12+17 -tab \ + lrgExtraTranscriptsHg38.bigPsl.bed ../../../chrom.sizes lrgBigPsl.bb + bigBedInfo lrgBigPsl.bb + ln -sf `pwd`/lrgBigPsl.bb /gbdb/hg38/bbi + hgBbiDbLink hg38 lrgBigPsl /gbdb/hg38/bbi/lrgBigPsl.bb + + # Load PSL, CDS and sequences. hgLoadPsl hg38 -table=lrgTranscriptAli lrgTranscriptsHg38.psl hgLoadSqlTab hg38 lrgCds ~/kent/src/hg/lib/cdsSpec.sql lrgTranscripts.cds hgPepPred hg38 tab lrgCdna lrgCdna.tab hgPepPred hg38 tab lrgPep lrgPep.tab ############################################################################# ## 7-Way Multiz (DONE - 2014-06-02 - Hiram) ssh hgwdev mkdir /hive/data/genomes/hg38/bed/multiz7way cd /hive/data/genomes/hg38/bed/multiz7way # from the 63-way in the source tree, select out the 7 used here: /cluster/bin/phast/tree_doctor \