This file was obtained following this steps:
Download from http://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/ the refGene.sql, knownGene.sql, knownToRefSeq.sql, kgXref.sql tables
Import everything it in mysql
Generate views for annotation:
SELECT kg.`chrom`, min(kg.cdsStart) as CDS_minStart,
max(kg.`cdsEnd`) as CDS_maxEnd, k2a.geneSymbol,
kg.`strand` as GeneStrand, min(kg.txStart) as TSS_minStart,
max(kg.txEnd) as TSS_maxStart,
kg.proteinID as ProteinID, k2a.protAcc as ProteinAcc, k2a.spDisplayID
FROM `knownGene` AS kg JOIN kgXref AS k2a
ON BINARY kg.name = k2a.kgID COLLATE latin1_bin
-- latin1_swedish_ci
-- WHERE k2a.spDisplayID IS NOT NULL and (k2a.`geneSymbol` LIKE 'Tcra%' or
k2a.`geneSymbol` LIKE 'TCRA%')
WHERE (k2a.spDisplayID IS NOT NULL or k2a.spDisplayID NOT LIKE '')
and k2a.`geneSymbol` LIKE 'Tcra%'
group by kg.`chrom`, k2a.geneSymbol
ORDER BY kg.chrom ASC , kg.txStart ASC