This is an implementation of the Compressively Accelerated Search Tools (CAST) in C99 built from the C implementation of CaBLASTP, which can be found on https://github.com/BurntSushi/c_cablastp. It is made to work like Po-Ru Loh's version of the code, which can be found in a prototype at http://cast.csail.mit.edu/.

The compressed databases created by this code are not in any way compatible with the compressed databases produced by the C++ implementation. Namely, the data itself should be the same (or similar), but its representation on disk is completely different. There may be other, subtler differences as well.

Multithreaded compression introduces some variance into the size of the resulting compressed database.

The official Go implementation of CaBLASTP can be found here: https://github.com/BurntSushi/cablastp

Notes about CaBLAT:

• Because of the smaller database size in CaBLAT, CaBLAT may produce hits that do not occur when running BLAT on a full uncompressed database.
• For CaBLAT, it is recommended to pass in the flags --number-queries and --number-targets to prevent errors in fine BLAT related to duplicate query and target sequence names.
  • To get a final .psl file in fine BLAT that contains the full names of the sequences in the query and target FASTA files and the locations of fine BLAT results in their original uncompressed sequences, pass in the option --complete-psl.
• Because of the differences between how fine BLAT is implemented in this version of CaBLAT and in Po-Ru's prototype version, there are some differences in the results of this version and the prototype.

CAST depends on three libraries: opt, ds and pthread. pthread should be installed via your system's package manager. opt and ds can be found in Andrew Gallant's [clibs respository] (https://github.com/BurntSushi/clibs), which is included in the clibs directory in src but is modified to produce .a files instead of .so files.

Additionally, to run CAST, blastn, makeblastdb and blat need to be installed in your PATH.

Briefly, the following commands should get CAST into a working state:

git clone git://github.com/BergerLab/CAST
cd CAST/src/clibs
make
cd ../..
make
./cablast-compress --help

The usage of cablast-compress is:

cablast-compress [flags] database-directory fasta-file [fasta-file ...]

• database-directory is the directory cablast-compress will create
• The FASTA file(s) being passed in are the database you want to compress.

The usage of cablast-search is:

./cablast-search [flags] database-dir query-fasta-file [ --blast_args BLASTN_ARGUMENTS ]

• query-fasta-file is a file containing the query sequences you wish to use for the CaBLAST search against a database previously compressed.
• --blast_args takes in any arguments to be passed directly to fine BLAST

We recommend that the query FASTA file not contain a large number of sequences because of a limitation with overlapping BLAST hits. We are working on fixing this limitation.

The usage of cablat is:

./cablat [flags] database-dir query-fasta-file results-file [ --blat_args BLAT_ARGUMENTS ]

• database-dir is the directory resulting from cablast-compress
• query-fasta-file is a file containing the query sequences you wish to use for the CaBLAST search against a database previously compressed.a
• results-file is the file that will contain the results of fine BLAT
• --blat_args takes in any arguments to be passed directly to fine BLAT