13.8 blastpgp Parameters (PSI-BLAST and PHI-BLAST)

blastpgp is the program used to run PSI-BLAST and PHI-BLAST. These programs are specialized protein BLAST comparisons that are more sensitive than the standard BLASTP search. PSI-BLAST considers position-specific information when searching for significant hits. PHI-BLAST uses a pattern, or profile, to seed an alignment, which is then extended by the normal BLASTP algorithm.

13.8.1 PSI-BLAST

PSI-BLAST (position-specific iterated BLAST) uses a specialized scoring matrix that assigns scores to each position (hence, position-specific) in the query sequence based on alignments defined by consecutive iterations of searches (hence, iterated). The specialized matrix is a position-specific scoring matrix (PSSM) that assigns a score for every amino acid at each position in the query sequence (See Figure 13-1).

Figure 13-1. PSSM for the first 10 amino acids of the coelacanth HoxA11 protein

Figure 13-1 shows a portion of a PSSM calculated for the coelacanth Hoxa11 protein (AAG39070). The query amino acids are numbered in the left column with the position-specific scores for each of the 20 amino acids shown across each row. The diverse scores of the three Tyrosines (Y) at positions 1, 7, and 8 highlight the position-specific aspect of this scoring scheme compared to traditional BLAST matrices, which would contain the same scores for Y in all three positions.

The PSSM, or checkpoint file, is created internally by PSI-BLAST, but it can also be exported to a file using the -C option of blastpgp. This option is extremely useful. You can use the checkpoint file in subsequent PSI-BLAST (blastpgp) searches or as a database entry for the RPS-BLAST program. You can also use the PSSM in a specialized tblastn search in blastall by using the -p psitblastn and -R <checkpoint file> options with a nucleotide database.

To run PSI-BLAST, the -j parameter must be set to something greater than 1. The default of -j 1 means that there are no iterations and that it's therefore the same as a single BLASTP search. Setting -j sets the maximum number of iterations to run, with the program stopping beforehand if the search comes to convergence. Convergence occurs when no new sequences are found that are better than the E value threshold set by the -h parameter.

Here are a few sample command lines:

blastpgp -d nr -i my_protein -s T -j 5
blastpgp -d nr -i my_protein -R my_protein.ckp -d nr -j 5 -h 0.001

13.8.2 PHI-BLAST

PHI-BLAST stands for pattern-hit initiated BLAST. The program uses an input sequence and a defined pattern to query a protein database. The pattern is defined in PROSITE format (http://ca.expasy.org/prosite/)and is used as the seed for the alignment. The pattern is used instead of the words that are usually generated for seeding alignments in BLASTP. Here's a sample profile:

ID  HoxA11 pattern1
PA  Y-S-[SA]-X-[LVIM]

The profile's syntax has a line starting with ID, followed by two spaces and the name of the pattern. The name is free text. The next line should start with PA, followed by two spaces, and then the pattern in PROSITE format. The PROSITE format is simple. A dash (-) separates letters, an X means any letter, and the brackets ([]) specify a choice of amino acids. You can find more information on the pattern syntax in the README.bls file that comes with the NCBI-BLAST distribution.

Additionally, if the pattern occurs more than once in the query and you would like to limit which occurrences are used as seeds, specify those locations by using the HI (hit initiation) tag in the pattern file. You set -p to seedp instead of patseedp (explained in the reference section that follows). The following example specifies that the pattern starting at position 143 should be used. (In this case, there's also an occurrence at 34, which is ignored.)

ID  HoxA11 pattern2
PA  Y-S-[SA]-X -[LVIMK]
HI  143

PHI-BLAST can also be a jumping-off point for a PSI-BLAST run. In the following command line, the pattern in hit_file initiates the first iteration of PSI-BLAST for the development of the PSSM, followed by normal rounds of PSI-BLAST iterations.

blastpgp -d nr -i my_protein -k hit_file -p patseedp -j 5

Here are a few sample PHI-BLAST command lines:

blastpgp -d nr -i my_protein -k hit_file -p patseedp
blastpgp -d nr -i my_protein -k multi_hit_file -p seedp
blastpgp -d HoxDB.pep -i AAG39070.pep -k hit_file.hox -p patseedp

The following reference describes parameters used with blastpgp, which executes PSI- and PHI-BLAST searches.

The number of processors to use; same as blastall.

The multiple-hit window size; same as blastall.

The number of alignments to show; same as blastall.

The input alignment file for a PSI-BLAST restart. It allows a PSI-BLAST run to start with a curated multiple sequence alignment instead of allowing the program to generate it from the first round of database alignments. For example:

blastpgp -i query -B multiple_alignment -j 5 -d nr

The alignment file must be based on the Clustal format but without the header and footer. The file should have a row for each sequence and can be broken into blocks separated by one or more blank lines. The query file (specified by -i) must be included in the alignment (though it doesn't need to be the first one), and all rows must be padded with dashes (—-) to make them equal lengths. Also, each column must contain either all uppercase or lowercase letters. An uppercase letter signifies that the column should be given a position-specific score; a lowercase letter means that the matrix (specified by -M) score should be used. Here is a portion of the example alignment file included in README.bls (the query is 26SPS9_Hs, in this case):

26SPS9_Hs     IHAAEEKDWKTAYSYFYEAFEGYdsidspkaitslkymllc
F57B9_Ce      LHAADEKDFKTAFSYFYEAFEGYdsvdekvsaltalkymll
YDL097c_Sc    ILHCEDKDYKTAFSYFFESFESYhnltthnsyekacqvlky
YMJ5_Ce       LYSAEERDYKTSFSYFYEAFEGFasigdkinatsalkymil
FUS6_ARATH    KNYIRTRDYCTTTKHIIHMCMNAilvsiemgqfthvtsyvn
COS41.8_Ci    SLDYKLKTYLTIARLYLEDEDPVqaemyinrasllqnetad
644879        KCYSRARDYCTSAKHVINMCLNVikvsvylqnwshvlsyvs
YPR108w_Sc    IHCLAVRNFKEAAKLLVDSLATFtsieltsyesiatyasvt
eif-3p110_Hs  SKAMKMGDWKTCHSFIINEKMNGkvw---------------
T23D8.4_Ce    SKAMLNGDWKKCQDYIVNDKMNQkvw---------------
YD95_Sp       IYLMSIRNFSGAADLLLDCMSTFsstellpyydvvryavis
KIAA0107_Hs   LYCVAIRDFKQAAELFLDTVSTFtsyelmdyktfvtytvyv
F49C12.8_Hs   LYRMSVRDFAGAADLFLEAVPTFgsyelmtyenlilytvit
Int-6_Mm      KFQYECGNYSGAAEYLYFFRVLVpatdrnalsslwgklase

26SPS9_Hs     kimlntpedvqalvsgklalryagrqtealkcvaqasknr
F57B9_Ce      ckvmldlpdevnsllsaklalkyngsdldamkaiaaaaqk
YDL097c_Sc    mllskimlnliddvknilnakytketyqsrgidamkavae
YMJ5_Ce       ckimlneteqlagllaakeivayqkspriiairsmadafr
FUS6_ARATH    kaeqnpetlepmvnaklrcasglahlelkkyklaarkfld
COS41.8_Ci    eqlqihykvcyarvldyrrkfleaaqrynelsyksaihet
644879        kaestpeiaeqrgerdsqtqailtklkcaaglaelaarky
YPR108w_Sc    glftlertdlkskvidspellslisttaalqsissltisl
eif-3p110_Hs  ----------------------------------------
T23D8.4_Ce    ----------------------------------------
YD95_Sp       gaisldrvdvktkivdspevlavlpqnesmssleacinsl
KIAA0107_Hs   smialerpdlrekvikgaeilevlhslpavrqylfslyec
F49C12.8_Hs   ttfaldrpdlrtkvircnevqeqltggglngtlipvreyl
Int-6_Mm      ilmqnwdaamedltrlketidnnsvssplqslqqrtwlih

Sets a constant in pseudocounts for PSSM. It's generally not necessary to change this parameter.

Outputs a file for PSI-BLAST checkpointing. This outputs the final PSSM for a multipass run of PSI-BLAST. The checkpoint file can then be used in a PSI-BLAST restart (see -R), in a blastall -p psitblastn run (also see -R), or as an entry in an RPS-BLAST database.

blastpgp -d nr -i my_protein -j 5 -C my_protein.ckp

The database name; same as blastall.

The expectation value; same as blastall.

The penalty to extend a gap; same as blastall.

The threshold for extending a hit; same as blastall.

Filters the query sequence; same as blastall.

Performs gapped alignment; same as blastall.

PHI-BLAST requires gapping and therefore forbids -g F.

The penalty to open a gap; same as blastall.

The E-value threshold for inclusion in PSSM. All alignments better than this threshold are used in constructing the PSSM.

The end of the required region in query. The default of -1 indicates the actual end of the query. This option can be used in combination with -S to specify a particular region to use

The query file; same as blastall.

Shows GIs in defline; same as blastall

The maximum number of passes to use in a multipass version. The default of 1 is just a regular BLASTP search.

Believes the query definition line; same as blastall.

Specifies the file containing the PROSITE pattern to be used for seeding in a PHI-BLAST run. If -k isn't specified when running PHI-BLAST (e.g. -p patseedp or -p seedp), the program looks for a file called hit_file.

The number of best hits from a region to keep; same as blastall.

Restricts the search of the database to a list of GIs; same as blastall.

The cost to decline an alignment.

Alignment view options; same as blastall.

The matrix; same as blastall.

The number of bits required to trigger gapping.

The output file for alignment; same as blastall.

A SeqAlign file output; same as blastall.

Specifies whether to run in PSI- or PHI-BLAST mode.

Options

blastpgp

PSI-BLAST mode

patseedp

PHI-BLAST mode. Uses all occurrences of the hit_file pattern to seed alignments. Any HI tags (see later) in the hit_file are ignored.

seedp

PHI-BLAST mode. The specified pattern is found more than once in the query, and the hit_file specifies which to use as seeds. The specific pattern(s) occurrences to use is specified with the HI tag in the hit_file. For example, the following hit_file designates seeding from a pattern that occurs at position 143 of the coelacanth HoxA11 protein:

ID  HoxA11 pattern2
PA  Y-S-[SA]-X-[LVIMK]
HI 143

seedp throws an exception if the hit_file doesn't contain the HI tags.

Output file for a PSI-BLAST matrix in ASCII format. This [file] can't be used in any subsequent programs. Use -c to output a matrix for subsequent searches.

Input checkpoint file for PSI-BLAST restart. Uses the checkpoint file. Output with -c.

Calculates locally optimal Smith-Waterman alignments. Because of the heuristic nature of BLAST, it sometimes produces nonoptimal local alignments. This option causes BLAST to run the full Smith-Waterman alignment algorithm on subjects found by the normal BLAST heuristic. There may be some speed cost using this option, but it helps guarantee high-quality alignments, which are important in PSSM generation. Setting -s T is highly recommended.

The start of the required region in query. Used in combination with -H, this sets a specific region of the query to be used when generating the PSSM.

Uses composition-based statistics. With this set to T, the score is adjusted based on composition biases in the query and subject sequences. Using it helps avoid possible corruption of the PSSM because it introduces low-entropy false positives in the multiple sequence alignment.

Produces HTML output; same as blastall.

Uses lowercase filtering of a query sequence; same as blastall.

The number of one-line descriptions to show; same as blastall.

The word size; same as blastall.

The X dropoff for gapped alignments; same as blastall.

X dropoff for ungapped extensions; same as blastall

The effective length of the search space; same as blastall.

The effective database size; same as blastall.

The X dropoff for final gapped alignment; same as blastall.