NAME¶

psi-cd-hit-2d-g1.pl - runs similar algorithm like CD-HIT but using BLAST to calculate similarities in db1 or db2 format

DESCRIPTION¶

Usage psi-cd-hit-2d [Options] Options

-i

in_dbname, required

-o

out_dbname, required

-c

clustering threshold (sequence identity), default 0.3

it means by default it doesn't use expect threshold, but with positive value, the program cluster seqs if similarities meet either identity threshold or expect threshold

-L

coverage of shorter sequence ( aligned / full), default 0.0

-M

coverage of longer sequence ( aligned / full), default 0.0

-R

(1/0) use psi-blast profile? default 0 perform psi-blast / pdb-blast type search

-G

(1/0) use global identity? default 1 sequence identity calculated as

total identical residues of local alignments / length of shorter seq

if you prefer to use -G 0, it is suggested that you also use -L, such as -L 0.8, to prevent very short matches.

-d

length of description line in the .clstr file, default 30 if set to 0, it takes the fasta defline and stops at first space

-l

length_of_throw_away_sequences, default 10

-p

profile search para, default

"-a 2 -d nr80 -j 3 -F F -e 0.001 -b 500 -v 500"

-s

blast search para, default

"-F F -e 0.000001 -b 100000 -v 100000"

-b

filename of list of hosts to run this program in parallel with ssh calls, you need provide a list of hosts

you can not use both ssh and pbs at same time

everytime after process 5000 sequences, program write a restart file and current clustering information

if program crash, stoped, termitated, you can restart it by add a option "-restart sth.restart"

if program clustered 200,000 seqs, it remove them from seq pool, and re format blast db to save time

when run in parallel with ssh (not pbs), I can copy blast dbs to local drives on each node to save blast db reading time BUT, IT MAY NOT FASTER

-J

job, job_file, exe specific jobs like parse blast outonly DON'T use it, it is only used by this program itself

so I won't run them as queries

means, that seq in db2 should <= seqs in db1 in a cluster

============================== by Weizhong Li, liwz@sdsc.edu ==============================

If you find cd-hit useful, please kindly cite:

"Clustering of highly homologous sequences to reduce thesize of large protein database", Weizhong Li, Lukasz Jaroszewski & Adam GodzikBioinformatics, (2001) 17:282-283 "Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences", Weizhong Li & Adam Godzik Bioinformatics, (2006) 22:1658-1659