dbiblast
Function
Description
dbiblast indexes a blast database created by the NCBI indexing programs
formatdb, pressdb or setdb, and builds EMBL CD-ROM format index
files. This format is used by the software on the EMBL database CD-ROM
distribution and by the Staden package in addition to EMBOSS, and
appears to be the most generally used and publicly available index
file format.
The index files of the blast database format are not useful by
themselves as they do not hold the entrynames or accession numbers,
but they are used as pointers to the reference and sequence information
in the database.
Because there are two sequence types (nucleic acid and protein) and two
index formats (blast1 from pressdb or setdb, and blast2 from formatdb)
with different index file names, dbiblast will ask about the database
type. If the answer is "unknown" dbiblast will test each possible set
of file names until one is found.
Usage
Command line arguments
Input file format
Any BLAST1 or BLAST2 database index files.
Output file format
dbiblast creates four index files. All are binary but with a
simple format.
- division.lkp is the master index file, and has a 300 byte header
containing the database name and date plus information on the record
size. This header is followed by one record for each database file,
giving the full file name for the data file, and optionally a second
sequence file.
- entryname.idx is the entry name index. It has the same 300 byte
header, mainly used to store the record size which will depend on the
size of the longest entryname in the database. Each entry is stored in
sorted alphanumeric order so that a binary search can be used to
efficiently find any record. The record also holds the file number
from division.lkp and the offsets in the data and sequence files for
that entry.
- acnum.trg holds the accession number information. The file has the
usual 300 byte header, and a sorted list of record by accession number.
Each accession number record contains the first record number in acnum.hit
and the total number of records in acnum.hit so that secondary (duplicated)
accession numbers can be searched.
- acnum.hit is a very simple file. After the usual 300 byte header,
each record simply holds the record number in entryname.idx. An
accession number search will use acnum.trg to find a start position
and number of records to read in this file, and will then simply
read the entryname.idx records for each entry in turn.
Data files
None.
Notes
dbiblast can't use the new style of Blast indices. You must
create the old style of Blast indices by adding -A F to the
formatdb command line.
Having created the EMBOSS indices for this file, a database can then be
defined in the file emboss.defaults as something like:
DB swnew [
type: P
format: embl
method: blast
directory: /data/embl
]
Fields Indexed
By default, dbiblast will index the ID name and the accession number (if present).
If they are present in your database, you may specify that
dbiblast should index the Sequence Version and GI number and the
words in the description by using the '-fields' qualifier with the
appropriate values.
Indexing the Blast database
dbiblast can't use the new style of Blast indices. You must
create the old style of Blast indices by adding -A F to the
formatdb command line.
The reason is that NCBI have not documented the new format, so we cannot
extend dbiblast to read it.
If you have the original (fasta format) file you can build the previous
blast database format with:
formatdb -A F
This turns off the new "ASN.1" formatting. blastall will still work.
References
None.
Warnings
Diagnostic Error Messages
None.
Exit status
None.
Known bugs
dbiblast can't use the new style of Blast indices. You must
create the old style of Blast indices by adding -A F to the
formatdb command line.
Author(s)
History
Target users
Comments