major version change: refactored _run_vep() to select consequence, variation, sift & polyphen values from specified gene/feature combination (multiple entries for same exon_id from vep, with varying sift, polyphen, consequence & variation vals taken into hashref meant last one wins); added cds and protein annotation vals; vep data row skipped & warning issued if ENSG0000* gene id not present in transcripts db
http://www.ensembl.org/info/docs/variation/vep/vep_script.html
Setup/update VEP - delete scripts/variant_effect_predictor dir
extract variant_effect_predictor.tar.gz into scripts dir
run scripts/variant_effect_predictor/INSTALL.pl - skip cache.
Cache manually installed/updated (needs to match VEP version) from:
ftp://ftp.ensembl.org/pub/release-<xx>/variation/VEP/
see README.txt in NGS/script dir for more details
Cache directory structure:
vep dir containing core & refseq dirs, each with homo_sapiens/xx cache
* on dev box - symlink /home/raj/.vep -> /media/sf_WIN_DRIVE/vep
* on deployment - /home/raj/.vep
to run from command-line:
perl script/variant_effect_predictor.pl -config=script/vep.ini \
-i=t/data/myeloid_variants.vep -o=output.txt --polyphen=b --sift=b \
--check_existing --coding_only --regulatory --dir=refseq/core
or:
perl script/variant_effect_predictor.pl -config=/home/raj/apps/NGS/script/vep.ini \
-o=output.txt --dir=/home/raj/.vep/refseq --polyphen=b --sift=b --check_existing \
--coding_only --regulatory -i=/tmp/20_06_14.vep
PERFORMANCE:
a) 4055-row vep input with unsorted chromosome order on 163.160.171.48:
fork=0 255 sec
fork=2 580 sec
fork=4 620 sec
b) 1028-row vep input with alpha-numerically sorted chromosome order on 163.160.171.48:
fork=0 95 sec
fork=2 65 sec (optimal)
fork=3 68 sec
fork=4 70 sec
fork=5 75 sec
fork=6 75 sec
performance much WORSE on dev server with fork enabled
splitting a 1110-row vep input into separate chromosome input files slightly
REDUCED performance (77sec vs 66sec) on dev
