[ Program Manual | User's Guide | Data Files | Databases ]
Table of Contents
Turning Hits Into HSPs
Generating Gapped Extensions


[Top | Next ]

BLAST searches one or more nucleic acid or protein databases for sequences similar to one or more query sequences of any type. BLAST can produce gapped alignments for the matches it finds.


[ Previous | Top | Next ]

BLAST, or Basic Local Alignment Search Tool, uses the method of Altschul et al. (J. Mol. Biol. 215: 403-410 (1990)) to search for similarities between a query sequence and all the sequences in a database.

This release of BLAST implements version 2 of BLAST from the National Center for Biotechnology Information (NCBI) described in Altschul et al. (Nucleic Acids Res. 25(17): 3389-3402 (1997)). BLAST is known as "gapped BLAST" because, in addition to offering a three-fold speedup over the original BLAST, it generates gapped alignments between query and database sequences.

You can specify any number of query sequences to BLAST, and they may be in any combination of protein or nucleic acid sequences. You can also specify any number of databases to BLAST, as long as all of the databases are of the same type (protein or nucleic acid). In the current release, if you want to specify multiple databases you must do so on the command line. In other words, you cannot specify more than one database from the interactive menu. For example:

% blast -INfile2=PIR,SWPLUS
You can also specify multiple queries using any valid multiple sequence specification. For example:
% blast -INfile1=hsp70.msf{*}
The GCG Wisconsin Package BLAST program supports five different programs in the BLAST family:
BLASTP, Protein Query Searching a Protein Database

Each database sequence is compared to each query in a separate protein-protein pairwise comparison.

BLASTX, Nucleotide Query Searching a Protein Database

Each query is translated, and each of the six products is compared to each database sequence in a separate protein-protein pairwise comparison.

BLASTN, Nucleotide Query Searching a Nucleotide Database

Each database sequence is compared to the query in a separate nucleotide-nucleotide pairwise comparison.

TBLASTN, Protein Query Searching a Nucleotide Database

Each nucleotide database sequence is translated, and each of the six products is compared to the queries in a separate protein-protein pairwise comparison.

TBLASTX, Nucleotide Query Searching a Nucleotide Database

The query and database sequences are translated in six frames, and each of the 12 products (for each query sequence) is compared in 36 different pairwise comparisons. Because this program involves more computation than the others, gapped alignments are not available when using TBLASTX.

Normally, BLAST decides which BLAST program you want to use simply by looking at the type (protein or nucleic acid) of your query sequence and the database you have selected. In the case of nucleotide-nucleotide searches, there are two programs that can do the search. By default, BLASTN is used. To search using TBLASTX instead, use -TBLASTX (but remember that gapped alignments are not available when using TBLASTX).

BLAST performs only local searches: It searches databases maintained at your institution. Local searches can consume significant computing resources, and require diligent maintenance of local databases. An alternative to running searches locally is to use NetBLAST which sends your query sequences over the internet to a server at NCBI, in Bethesda, MD. Keep in mind, however, that NCBI imposes some limititions on NetBLAST searches such as restricting the number of searches that a user is permitted to run in a single day, and prohibiting TBLASTX searches against the NR database. Additionally, NetBLAST does not support as many search options as are available with BLAST.

BLAST is a statistically driven search method that finds regions of similarity between your query and database sequences and produces gapped alignments of these regions. Within these aligned regions, the sum of the scoring matrix values of their constituent symbol pairs is higher than some level that you would expect to occur by chance alone.

You are prompted to set an expectation level for the entire search. The expectation of a sequence is the probability of the current search finding a sequence with as good a score by chance alone. Therefore setting the maximum expectation level to 10.0, the default, limits the reported sequences to those with scores high enough to be have been found by chance only ten or fewer times.


[ Previous | Top | Next ]

Here is a session using BLAST to find the sequences in PIR with similarities to a myoglobin gene:

% blast

 BLAST with what query sequence(s) ?  mywhp.pep

                  Begin (* 1 *) ?
                End (*   153 *) ?

 Search for query in what sequence database:

   1) pir     p Protein Information Resource
   2) swplus  p SWISS-PROT+SP-TREMBL
   3) genembl n GenBank+EMBL
   4) est     n Expressed Sequence Tags
   5) sts     n Sequence Tagged Sites
   6) gss     n Genome Survey Sequences

 Please choose one (* 1 *):

 Ignore hits expected to occur by chance more than (* 10.0 *) times?

 Limit the number of sequences in my output to (* 500 *) ?

 What should I call the output file (* mywhp.blastp *) ?

 1 Searching pir with query pir1:mywhp...done.

   CPU time (sec): 10.4
      Output file: mywhp.blastp

 Number of query sequences searched: 1
                     CPU time (sec): 10.4



[ Previous | Top | Next ]

Below is part of the output from the search in the example session:

The output has four parts: 1) an introduction that tells where the search occurred and what database and query were compared; 2) a list of the sequences in the database containing HSPs (high-scoring segment pairs) whose scores were least likely to have occurred by chance (the entries in this list have begin and end ranges on them if -FRAGments is specified); 3) a display of the alignments of the HSPs showing identical and similar residues; and 4) a complete list of the parameter settings used for the search.

By default, BLAST looks for alignments that contain gaps. If you only look for alignments that do not contain gaps, there will often be more than one segment pair associated with each database sequence.


BLASTP 2.0.5 [May-5-1998]

Reference: Altschul, Stephen F., Thomas L. Madden, Alejandro A. Schaffer,
Jinghui Zhang, Zheng Zhang, Webb Miller, and David J. Lipman (1997),
"Gapped BLAST and PSI-BLAST: a new generation of protein database search
programs",  Nucleic Acids Res. 25:3389-3402.
         (153 letters)

Database: pir
           107,076 sequences; 34,138,851 total letters

Searching. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
                                                                   Score     E
Sequences producing significant alignments:                        (bits)  Value
PIR1:MYWHP  Begin: 1 End: 153 !myoglobin - sperm whale                313  1e-85
PIR1:MYWHW  Begin: 1 End: 153 !myoglobin - dwarf sperm whale          305  3e-83

PIR1:MYSLG  Begin: 2 End: 153 !myoglobin - gray seal                  270  1e-72

PIR1:MYAQ  Begin: 3 End: 154 !myoglobin - American alligator          206  2e-53
\\End of List
>PIR1:MYWHP myoglobin - sperm whale
           Length = 153

 Score =  313 bits (794), Expect = 1e-85
 Identities = 153/153 (100%), Positives = 153/153 (100%)




>PIR1:MYWHW myoglobin - dwarf sperm whale
           Length = 153

 Score =  305 bits (773), Expect = 3e-83
 Identities = 148/153 (96%), Positives = 151/153 (97%)




>PIR1:MYSLG myoglobin - gray seal
           Length = 153

 Score =  270 bits (682), Expect = 1e-72
 Identities = 127/152 (83%), Positives = 140/152 (91%)




>PIR1:MYAQ myoglobin - American alligator
           Length = 154

 Score =  206 bits (519), Expect = 2e-53
 Identities = 94/152 (61%), Positives = 122/152 (79%)

           LS+ EW+ VL +W KVE+ +  HG +++IRL + HPET E+F++FKH+KT  EMK+SE +



  Database: pir
    Posted date:  Jul 8, 1998  4:54 PM
  Number of letters in database: 34,138,851
  Number of sequences in database:  107,076

Lambda     K      H
   0.318    0.135    0.392

Lambda     K      H
   0.270   0.0470    0.230

Matrix: BLOSUM62
Gap Penalties: Existence: 11, Extension: 1
Number of Hits to DB: 10131438
Number of Sequences: 107076
Number of extensions: 379101
Number of successful extensions: 1983
Number of sequences better than 10: 645
Number of HSP's better than 10.0 without gapping: 572
Number of HSP's successfully gapped in prelim test: 73
Number of HSP's that attempted gapping in prelim test: 1152
Number of HSP's gapped (non-prelim): 652
length of query: 153
length of database: 34138851
effective HSP length: 50
effective length of query: 103
effective length of database: 28785051
effective search space: 2964860253
T: 11
A: 40
X1: 16 ( 7.3 bits)
X2: 38 (14.8 bits)
X3: 64 (24.9 bits)
S1: 41 (21.7 bits)
S2: 61 (28.2 bits)
The BLAST output is a list file that is suitable for input to any GCG program that allows indirect file specifications. For information about indirect file specification, see Chapter 2 of the User's Guide, Using Sequence Files and Databases.


[ Previous | Top | Next ]
Bit Score

Each aligned segment pair has a normalized score expressed in bits that lets you estimate the magnitude of the search space you would have to look through before you would expect to find an HSP score as good as or better than this one by chance. If the bit score is 30, you would have to score, on average, about 1 billion independent segment pairs (2(30)) to find a score this good by chance. Each additional bit doubles the size of the search space. This bit score represents a probability; one over two raised to this power is the probability of finding such a segment by chance. Bit scores represent a probability level for sequence comparisons that is independent of the size of the search.

The size of the search space is proportional to the product of the query sequence length times the sum of the lengths of the sequences in the database. This product, referred to as N in Altschul's publications, is multiplied by a coefficient K to get the size of the search space. When searching protein databases with protein queries, K is about 0.13. BLAST uses estimates of K produced before it runs by random simulation (Altschul & Gish, Methods in Enzymology 266; 460-480 (1996)).

E Value

There is a probability associated with each pairwise comparison in the list and with each segment pair alignment. The number shown in the list is the probability that you would observe a score or group of scores as high as the observed score purely by chance when you do a search against a database of this size.

An ideal search would find hits that go from extremely unlikely to ones whose best scores should have occurred by chance alone (that is, with probabilities approaching 1.0).


If you specify ungapped alignments to BLAST, a third column of data will appear in your output under the heading N. The number in that column indicates how many HSPs were involved in computing the statistics for the sequence. If the number is greater than 1, the scores of multiple HSPs were combined to produce the result. See the ALGORITHM topic for more information.

BLAST Parameters

At the end of the output is a listing of parameter settings along with some trace information about the search. Some of these parameters are described in this document, but to get more complete documentation on these parameters, look at the BLAST release notes on the World Wide Web at


[ Previous | Top | Next ]

BLAST accepts any number of protein or nucleic acid sequences as input. The search set is a specially formatted database. See the GCGToBLAST entry in the Program Manual for information on how to create a local database that BLAST can search from a set of sequences in GCG format.

The function of BLAST depends on whether your input sequence(s) are protein or nucleotide. Programs determine the type of a sequence by the presence of either Type: N or Type: P on the last line of the text heading just above the sequence. If your sequence(s) are not the correct type, turn to Appendix VI for information on how to change or set the type of a sequence.


[ Previous | Top | Next ]

PSIBLAST iteratively searches one or more protein databases for sequences similar to one or more protein query sequences. PSIBLAST is similar to BLAST except that it uses position-specific scoring matrices derived during the search.

NetBLAST searches for sequences similar to a query sequence. The query and the database searched can be either peptide or nucleic acid in any combination. NetBLAST can search only databases maintained at the National Center for Biotechnology Information (NCBI) in Bethesda, Maryland, USA.

GCGToBLAST combines any set of GCG sequences into a database that you can search with BLAST.

FastA does a Pearson and Lipman search for similarity between a query sequence and a group of sequences of the same type (nucleic acid or protein). For nucleotide searches, FastA may be more sensitive than BLAST.

TFastA does a Pearson and Lipman search for similarity between a protein query sequence and any group of nucleotide sequences. TFastA translates the nucleotide sequences in all six reading frames before performing the comparison. It is designed to answer the question, "What implied protein sequences in a nucleotide sequence database are similar to my protein sequence?"

FastX does a Pearson and Lipman search for similarity between a nucleotide query sequence and a group of protein sequences, taking frameshifts into account. FastX translates both strands of the nucleic sequence before performing the comparison. It is designed to answer the question, "What implied protein sequences in my nucleic acid sequence are similar to sequences in a protein database?"

TFastX does a Pearson and Lipman search for similarity between a protein query sequence and any group of nucleotide sequences, taking frameshifts into account. It is designed to be a replacement for TFastA, and like TFastA, it is designed to answer the question, "What implied protein sequences in a nucleotide sequence database are similar to my protein sequence?"

SSearch does a rigorous Smith-Waterman search for similarity between a query sequence and a group of sequences of the same type (nucleic acid or protein). This may be the most sensitive method available for similarity searches. Compared to BLAST and FastA, it can be very slow.

FrameSearch searches a group of protein sequences for similarity to one or more nucleotide query sequences, or searches a group of nucleotide sequences for similarity to one or more protein query sequences. For each sequence comparison, the program finds an optimal alignment between the protein sequence and all possible codons on each strand of the nucleotide sequence. Optimal alignments may include reading frame shifts.

WordSearch identifies sequences in the database that share large numbers of common words in the same register of comparison with your query sequence. The output of WordSearch can be displayed with Segments.

ProfileSearch and MotifSearch use a profile (derived from a set of aligned sequences) instead of a query sequence to search a collection of sequences.

HmmerSearch uses a profile hidden Markov model as a query to search a sequence database to find sequences similar to the family from which the profile HMM was built. Profile HMMs can be created using HmmerBuild.

FindPatterns uses a pattern described by a regular expression to search a collection of sequences. Motifs looks for sequence motifs by searching through proteins for the patterns defined in the PROSITE Dictionary of Protein Sites and Patterns. Motifs can display an abstract of the current literature on each of the motifs it finds.


[ Previous | Top | Next ]

Because of the way BLAST must estimate certain statistical parameters (see the ALGORITHM topic elsewhere in this document), the number of scoring matrices available for use with BLAST is limited. Currently, valid choices for the -MATRix parameter are BLOSUM62 (the default), BLOSUM45, BLOSUM80, PAM30, and PAM70.

Gap creation and gap extension penalties are supported in limited combinations depending upon which scoring matrix is in use. The following table shows the allowed combinations for amino acids. The first values listed are the defaults for each scoring matrix.


 Scoring Matrix    Gap Opening Penalty    Gap Extension Penalty 

   BLOSUM62                 11                       1 
                             7                       2                              8                       2                              9                       2                             10                       1                             12                       1
   BLOSUM80                 10                       1                               6                       2                              7                       2                              8                       2                              9                       1                             11                       1
   BLOSUM45                 14                       2                                   10                       3                             11                       3                             12                       3                             13                       3                             12                       2                             13                       2                             15                       2                             16                       1                             17                       1                             18                       1                             19                       1
    PAM30                    9                       1                               5                       2                              6                       2                              7                       2                              8                       1                             10                       1
    PAM70                   10                       1                               6                       2                              7                       2                              8                       2                              9                       1                             11                       1

Gapped alignments are not an option when running TBLASTX.

You may choose multiple query sequences, any of which may be either nucleic acid or protein. You may also choose multiple databases against which to search, however each of these must be of the same type.

If you used GCGToBLAST to create your BLAST databases from any source other than a GCG-formatted database (such as from arbitrary sequence files, an MSF or RSF file, etc.), then BLAST's list file output won't be a functional list file. If you want to take full advantage of BLAST's list file output, make sure that you generate your BLAST databases from a GCG-formatted database. You can use DataSet to generate such databases from any set of sequences in GCG format.


[ Previous | Top | Next ]

BLAST can search only a specially compressed form of the data. Therefore, you can search only those databases that are available in this form, and you must search them in their entirety. If you want to restrict the search to a specific set of sequences, use the program GCGToBLAST to create a specially compressed database consisting of just those sequences.

To name a searchable database interactively, choose the number of the database of interest from the menu. Use a parameter like -INfile2=genbank to choose the name of the database you want to search.

If a nucleic acid and a protein database share the same name, BLAST cannot be sure which one of them you mean when you specify one of them using the -INfile2 parameter. If the database you want to search cannot be named unambiguously with the -INfile2 parameter, add either -DBNucleotideonly or -DBProteinonly to the command line.


[ Previous | Top | Next ]

BLAST is a client for an implementation of gapped BLAST (Altschul et al., Nucleic Acids Research 25; 3389-3402 (1997)), an heuristic algorithm for searching protein and nucleic acid databases for similarities to query sequences.

The above example demonstrates BLASTP, which searches for similarities between protein queries and protein databases, as a prototype for BLAST. However, the ideas are immediately applicable to comparisons involving conceptual translations of query sequences and databases, and extend to similarity searches between nucleic acid sequences as well.

BLAST compares a query sequence with a database sequence by first locating two non-overlapping sequence segments in common within a certain distance of each other, and then attempts to extend these putative "hits" into locally optimal alignments between the sequences being compared. A more detailed description is provided below.


[ Previous | Top | Next ]

BLAST uses a substitution matrix (such as the BLOSUM or PAM matrices) to assign a score to the alignment of any pair of amino acids. An aggregate score for an alignment segment can be computed by summing the scores of each amino acid pair in that segment. When given two sequences to compare, the original (ungapped) BLAST algorithm searches for arbitrary but equal length segments within each sequence that have a maximal aggregate score which meets or exceeds some threshold or cutoff score. BLAST looks for locally optimal alignments between the two sequences whose scores cannot be improved either by extending or trimming. Such locally optimal alignments are called "high-scoring segment pairs," or HSPs.

If you assume a simple protein model in which amino acids occur randomly at all positions and in proportion to the frequencies at which they are found within the database and query sequences, then we can compute a normalized score (expressed in units called bits) from the nominal score of an HSP. Such normalized scores allow direct statistical comparison of results regardless of the scoring system used (see "Generating Gapped Extensions" for a caveat to this). Furthermore, the normalized score can be used to compute an expect value, or E-value, which is the number of distinct HSPs having at least that normalized score expected to occur by chance. This theory has not been proved for gapped local alignments and their associated scores, but there are indications that it remains valid (Altschul et al., 1997).

Turning Hits Into HSPs

[ Previous | Top | Next ]

The central idea of the BLAST algorithm is that any statistically significant alignment between two sequences is likely to contain a high-scoring pair of aligned words. A word is simply a sequence segment of specified length (usually 3 for protein sequences). BLAST begins its comparison of a query sequence to a database by scanning the database for words that score at least the threshold score T when aligned with some word within the query sequence. Any word pair satisfying this condition is called a hit. The diagonal of a hit involving words starting at positions (x, y) of the database and query sequences is defined as x-y. The distance between two hits on the same diagonal is defined as the difference between their first coordinates.

Once a hit is found, BLAST determines whether the hit lies within an alignment having an aggregate score high enough to be reported. It does this by extending the hit in both directions until the running alignment's score has dropped more than some quantity X below the maximum score yet attained. This extension step is quite costly, taking upwards of 90% of BLAST's execution time under most circumstances.

In order to reduce the number of extensions it has to perform, BLAST takes advantage of the fact that an interesting HSP is typically much longer than a single hit. In fact, it is likely to contain multiple hits on the same diagonal within a relatively short distance of one another. Therefore, BLAST chooses a length A and invokes an ungapped extension if and only if two non-overlapping hits are found on the same diagonal within distance A of one another. (Any hit that overlaps the most recent one is ignored.)

Generating Gapped Extensions

[ Previous | Top | Next ]

Gapped extensions allow BLAST to maintain its sensitivity while tolerating a much higher chance of missing any single moderately scoring HSP. However, gapped extensions take about 500 times longer to execute than ungapped extensions. Therefore, BLAST triggers a gapped extension for an HSP only when its score exceeds a moderate score (Sg) specifically chosen so that no more than about one gapped extension is invoked per 50 database sequences.

To generate the gapped local alignment, BLAST uses a standard dynamic programming algorithm for pairwise sequence alignment which traverses the cells of a path graph, the dimensions of which are the lengths of the two sequences being compared, performing a fixed amount of computation per each cell. Starting from a single aligned pair of residues, called the seed, the dynamic programming proceeds both forward and backward through the path graph considering only those cells for which the optimal local alignment score falls no more than X below the best alignment score yet found. (This description is a generalization of BLAST's method for constructing HSPs.) The region of the path graph explored adapts to the alignment being produced.

The seed for the dynamic programming is the central residue pair of the length-11 segment of the HSP having the highest alignment score. If the HSP itself is shorter than 11 residues in length, its central pair of residues is chosen.

The resulting gapped alignment is reported only if it has an E-value low enough to be of interest. For any alignment actually reported, BLAST performs a gapped extension that records "traceback" information (Sankoff and Kruskal, 1983) using a substantially larger X parameter than that employed during the search stage to increase the accuracy of the alignment.

Because BLAST produces gapped alignments only for those few database sequences likely to be related to the query, it cannot estimate the parameters necessary to compute normalized scores on the fly. Instead, BLAST must rely on estimates of these parameters generated beforehand by random simulation. For this reason, BLAST cannot use a scoring system for which no simulation has been performed and still produce accurate estimates of statistical significance.


[ Previous | Top | Next ]
Bit Scores and the Size of the Search

Altschul has shown that for sequences that have diverged by a certain amount, there is an informativeness (or ability to discriminate between chance scores and significant scores) associated with each residue pair in the segment pair. This informativeness is the amount of information obtainable from each residue pair in a real alignment that can be used to distinguish the real alignment from a random one. This informativeness can be expressed in bits. The sum of the information available from each residue pair in a segment is the segment pair's score in bits. Such scores are intuitively understandable as the significance of a segment pair score. To express such scores as a fraction you would divide 1 by 2 to the number of bits in the score. For example, if a segment pair has a bit-score of 16, then the appropriate fraction (1/2(16)=1/65,536) would suggest that you should see a score this high by chance about once for every 65,000 independent segment pairs you examine.

For nucleotide sequences that have not diverged, there should be an informativeness of about 2 bits per nucleotide pair. For protein sequences that have not diverged, the informativeness should be slightly over 4 bits per amino acid pair. (The informativeness per pair goes down as the sequences diverge and a segment pair score is maximally informative only when a scoring matrix appropriate to the extent of divergence between the sequences is used to calculate the score.)

The bit scores are absolute, but the expectation of finding any particular score depends on the size of the search space. The number of places where a segment pair might originate is proportional to the product of the length of the query times the sum of the lengths of all the sequences searched. This product is multiplied by a coefficient K to get the size of the search space. When searching protein databases with protein queries, K is approximately 0.13.

For a query sequence of length 300 aa searching a database of 12 million residues, the size of the search space would be 300 x 12,000,000 x 0.13 or 468,000,000. For a search this size, a score that only occurs once in every 65,000 potential segment pairs (that is, with a bit score of 16) would be expected to occur about 7,200 times by chance alone.

If the database being searched is highly redundant (as it might be if it contained several hundred homologous cytochromes), then size of the search space calculated by these methods will overestimate the size of the real search space.

Using BLAST for Nucleotide Searches

The detection of distant relationships between proteins is easier than between nucleotide sequences, even if the nucleotide sequences have to be translated in all six frames to make the amino acid comparison. To give a rough magnitude to this generalization, it is possible to detect similarities in proteins that have diverged by 250 substitutions per 100 residues (250 PAM units) while nucleotide similarities become obscure at distances much greater than 50 substitutions per 100 nucleotides (50 DNA PAM units). Nonetheless, when the nucleotide sequences being compared do not code for proteins, you have no alternative but to search at the nucleotide level. We suggest you consider either reducing the word size for BLAST from its default of 11 to perhaps six or seven, or using the FastA program when looking for nucleotide homologs.

Increasing Program Speed Using Multithreading

This program is multithreaded. It has the potential to run faster on a machine equipped with multiple processors because different parts of the analysis can be run in parallel on different processors. By default, the program assumes you have one processor, so the analysis is performed using one thread. You can use -PROCessors to increase the number of threads up to the number of physical processors on the computer.

Under ideal conditions, the increase in speed is roughly linear with the number of processors used. But conditions are rarely ideal. If your computer is heavily used, competition for the processors can reduce the program's performance. In such an environment, try to run multithreaded programs during times when the load on the system is light.

As the number of threads increases, the amount of memory required increases substantially. You may need to ask your system administrator to increase the memory quota for your account if you want to use more than two threads.

Never use -PROCessors to set the number of threads higher than the number of physical processors that the machine has -- it does not increase program performance, but instead uses up a lot of memory needlessly and makes it harder for other users on the system to get processor time. Ask your system administrator how many processors your computer has if you aren't sure.

When Blastall Produces No Output

You may see an error indicating that blastall produced no output (blastall is the name of the BLAST executable provided by NCBI). One of the possible causes of this condition is the presence of a file in your home directory called ".ncbirc" which contains an invalid path to the NCBI data directory. The NCBI data directory should contain seqcode.val, gc.code, BLOSUM62, and perhaps some other data files. If your home directory does indeed contain such a file, we recommend that you either rename it (the safest option), edit it to update the path to the NCBI data directory (this takes some effort, but that path is contained in the logical name "NCBI"), or delete it (the simplest option). Your system administrator should be able to help you do this if you have trouble, or you may contact support at


When searching a nucleotide database with a protein query (i.e. when using TBLASTN) you may optionally use a position-specific matrix (PSSM) instead of a standard scoring matrix. This kind of search is called PSI-TBLASTN and it is enabled when you use -REStorecheckpoint to specify a checkpoint file that was created in advance using the program PSIBLAST.

A checkpoint file contains both the PSSM and the query that was used when running PSIBLAST. For this reason, when performing a PSI-TBLASTN search, you must use the exact same query sequence that was used when the checkpoint file was saved. In addition, checkpoint files are platform-specific binary files which means that checkpoint files created with PSIBLAST one operating system will not work correctly when running BLAST on a different type of system.

BLAST filters query sequences by default, in contrast to PSIBLAST which does not. For the sake of compatibility, when you plan to use a PSSM from PSIBLAST to perform a PSI-TBLASTN search you should specify -NOFILter unless you specified -FILter when you ran PSIBLAST.


[ Previous | Top | Next ]
List Size Limit

A list size that is too small to display all the significant hits is a common problem. To see the unlisted hits you must run the search again with the list size limit set high enough to include everything significant.

Segment Pair Alignment Limit

BLAST displays alignments of segment pairs from the top 250 sequences in the list. You can adjust this limit with -ALIgnments. BLAST will not show alignments for sequences not present in the list.


For nucleotide sequence comparisons, the word size defaults to 11 -- no segment pair can be scored unless it contains a perfect match of at least 11 consecutive bases. If sensitivity is much more important than selectivity, and your search cannot be done at the amino acid level, you might want to reduce the word size to seven or even six. NCBI has stated that there is only a marginal increase in sensitivity for settings smaller than this.

BLAST uses a word size of three for proteins (11 for blastn searches), which is appropriate for a wide range of searches, but you can adjust the synonym threshold T downwards to increase sensitivity at the price of speed. Read the PARAMETER REFERENCE topic for more information on -HITEXTTHRESHold and -EXPect.

Batch Queue

Using BLAST to search a large local database can take a long time. You may want to run searches in the batch queue. You can specify that this program run at a later time in the batch queue by using -BATch. Run this way, the program prompts you for all the required parameters and then automatically submits itself to the batch or at queue. For more information, see "Using the Batch Queue" in Chapter 3, Using Programs in the User's Guide.

Relationship to FastA

For protein database searches, BLAST and FastA have similar sensitivity, although the different algorithms employed make it possible, at least in principle, for FastA to find things that BLAST misses and vice versa. For nucleotide database searches with nucleotide query sequences, FastA may be more sensitive, since by default BLAST ignores segment pairs that do not contain a perfect match of at least 11 adjacent nucleotides (22 bits). This default misses many obviously significant relationships. If you are looking for nucleotide sequence homologs that do not code for proteins (that is, if your search cannot be done at the amino acid level), we suggest you either reduce the word size to seven or use the FastA program instead of BLAST.


[ Previous | Top | Next ]

BLAST filters out regions of low complexity from query sequences by default. You can turn filtering off by using the -NOFILter parameter. Searches against a nucleotide database with nucleotide queries (blastn) employ the DUST filter program (Hancock and Armstrong, Comput. Appl. Biosci. 10: 67-70 (1994); Tatusov and Lipman, unpublished). All other searches employ the SEG filter program (Wootton and Federhen, Computers in Chemistry 17: 149-163 (1993); Wootton and Federhen, Methods in Enzymology 266: 554-571 (1996)). For a general discussion of the role of filtering in search strategies, see Altschul et al., Nature Genetics 6: 119-129 (1994).

Short repeats and low complexity sequences, such as glutamine-rich regions, confound most database searching methods. For BLAST, the random model against which the significance of segment pair scores is evaluated assumes that at each position, each residue has a probability of occurring which is proportional to its composition in the database as a whole. Low complexity or highly repetitive sequences are inconsistent with this assumption.

Low complexity sequence found by the filter program is substituted using the letter N in nucleotide sequence and the letter X in amino acid sequence. Here is an example of a sequence aligned to a filtered copy of itself to show which parts are filtered out:






[ Previous | Top | Next ]

BLAST normally uses the BLOSUM62 scoring matrix from Henikoff and Henikoff (Proc. Natl. Acad. Sci. USA 89; 10915-10919 (1992)) whenever the sequences being compared are proteins (including cases where nucleotide databases or query sequences are translated into protein sequences before comparison). You can use other BLOSUM45, BLOSUM80, or the more traditional PAM70 and PAM30 scoring matrices with -MATrix, for example -MATrix=PAM40. Each matrix is most sensitive for finding homologs at the corresponding PAM distance. The seminal paper on this subject is Stephen Altschul's "Amino acid substitution matrices from an information theoretic perspective" (J. Mol. Biol. 219; 555-565 (1991)). If you are new to this literature, an easier place to start reading might be Altschul et al., "Issues in searching molecular sequence databases" (Nature Genetics, 6; 119-129 (1994)).


[ Previous | Top | Next ]

There is no external scoring matrix for nucleotide-nucleotide searches (that is, searches where both the query and the database are nucleotide sequences and where you have not used -TBLASTX. But as is explained below you can specify a nucleotide-nucleotide scoring matrix for any PAM distance by changing the match/mismatch ratio. The default ratio is +1/-3. You can change the ratio by specifying a new value for the numerator using -MATCH.


[ Previous | Top | Next ]

BLAST normally uses the standard genetic code if either the query or the database sequences requires translation. If your query comes from a system where this genetic code is inappropriate, you can select any of these alternative codes by the numbers given in the following table:

     1 Standard or Universal
     2 Vertebrate Mitochondrial
     3 Yeast Mitochondrial
     4 Mold, Protozoan, Coelenterate Mitochondrial and Mycoplasma/Spiroplasma
     5 Invertebrate Mitochondrial
     6 Ciliate Macronuclear
     7 [Do not use this index]
     8 [Do not use this index]
     9 Echinodermate Mitochondrial
    10 Alternative Ciliate(Euplotid) Macronuclear
    11 Eubacterial
    12 Alternative Yeast
    13 Ascidian Mitochondrial
    14 Flatworm Mitochondrial
    15 Alternate Ciliate (Blepharisma) Nuclear
    16 Chlorophycean Mitochondrial
    21 Trematode Mitochondrial
You can specify the genetic code for the query and the database independently. Use -TRANSlate=2 to tell BLAST to use the vertebrate mitochondrial code to translate the query. Use -DBTRANSlate=3 to tell BLAST to use the yeast mitochondrial code to translate the database. (Note that most of the genes in GenBank will be translated inappropriately if you select a nonstandard genetic code for database translation.)


[ Previous | Top | Next ]

BLAST searches only local databases. See the NetBlast entry in the Program Manual for information on how to run BLAST searches remotely.


[ Previous | Top | Next ]

All parameters for this program may be added to the command line. Use -CHEck to view the summary below and to specify parameters before the program executes. In the summary below, the capitalized letters in the parameter names are the letters that you must type in order to use the parameter. Square brackets ([ and ]) enclose parameter values that are optional. For more information, see "Using Program Parameters" in Chapter 3, Using Programs in the User's Guide.

Minimal Syntax: % blast [-INfile1=]pir:mywhp  -Default

Prompted Parameters:

-BEGin=1 -END=153        sets the ranges of interest in  query sequences
[-INfile2=]pir           specifies database(s) to search
-EXPect=10.0             ignores scores that would occur by chance
                           more than 10 times
-LIStsize=500            sets maximum number of sequences listed in the output
[-OUTfile=]mywhp.blastp  names the output file

Local Data Files:

[-DATa2=blast.ldbs]      names the list of available local databases
[-DATa3=blast.sdbs]      names the list of available site-specific databases

Optional Parameters:

-PROCessors=1            sets the number of processors to use
-TBLASTX                 if query and database are both nucleotide,
                           translates both and does protein comparisons
-DBNucleotideonly        searches only nucleic databases
-DBProteinonly           searches only protein databases
-WORdsize=0              sets word size (0 selects program default)
-MATch=1                 sets nucleotide match reward
-MISmatch=-3             sets nucleotide mismatch penalty
-MATRix=blosum62         assigns the scoring matrix for proteins
-GAPweight=0             sets gap creation penalty
-LENgthweight=0          sets gap extension penalty
-HITEXTTHRESHold=0       sets minimum score to extend hits
-NOFILter                suppresses filtering of low complexity segments
                           out of nucleotide and protein query sequences
-TRANSlate=1             names genetic code for translating query
-DBTRANSlate=1           names genetic code for translating database
-EFFdbsize=0             sets effective database size (0 real size)
-NOFRAgments             suppresses showing list file entries as fragments
-ALIgnments=250          sets number of sequences for which to show
-VIEW=0                  selects alignment view type (0-8 allowed)
-NOGAPS                  suppresses gapped alignments
-XDRopoff=0              sets X dropoff value for gapped alignments (X2)
-MEGAblast               uses MegaBLAST algorithm for search
-REStorecheckpoint[=mywp.chk] reads checkpoint file and runs PSI-TBLASTN
-LOWercasemask           filters lower case characters in query sequence
-HITWindow=40            sets multiple hist window size (A)
-BESthits                sets number of best hits from a region to keep (K)
-HTML                    uses HTML for output format
-NATive                  produces unmodified BLAST2 output
-APPend="string"         appends "string" to pass-through command line
-BATch                   submits program to batch queue
-DBReport                lists valid databases then exits


[ Previous | Top | Next ]

The original paper describing BLAST is Altschul, Stephen F., Gish, Warren, Miller, Webb, Myers, Eugene W., and Lipman, David J. (1990). Basic local alignment search tool. J. Mol. Biol. 215; 403-410. Gapped BLAST is described in Altschul, Stephen F., Madden, Thomas L., Schaffer, Alejandro A., Zhang, Jinghui, Zhang, Zheng, Miller, Webb, and Lipman, David J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25(17); 3389-3402.


[ Previous | Top | Next ]

BLAST was written by Warren Gish, formerly of the National Center for Biotechnology Information (NCBI), in collaboration with Stephen Altschul, Webb Miller, Eugene Myers, David Lipman, and David States. The document you are now reading was written by John Devereux, with some modifications for BLAST2 by Ted Slater.

Blastall (NCBI's implementation of BLAST 2.0) was written for NCBI by Tom Madden. GCG's client for blastall was written by Ted Slater for distribution with the Wisconsin Package Version 10.0; some portions were taken from GCG's original BLAST client written by Scott Rose. The output post-processor for release 10.0 was written by Ron Stewart.

We are extremely grateful to Stephen Altschul and Warren Gish for their careful and original work on BLAST and for their critical comments on GCG's BLAST documentation, and we are very grateful to NCBI for making these programs and services available to the molecular biology community.


[ Previous | Top | Next ]

The files described below supply auxiliary data to this program. The program automatically reads them from a public data directory unless you either 1) have a data file with exactly the same name in your current working directory; or 2) name a file on the command line with an expression like -DATa1=myfile.dat. For more information see Chapter 4, Using Data Files in the User's Guide.

BLAST reads two files, blast.ldbs (local databases), and blast.sdbs (site-specific databases). These together list the search sets in the menu. We update blast.ldbs when we send database updates to your institution. If you have sequences of local interest that you would like to search with BLAST, read the documentation for GCGToBLAST to see how to create local BLAST-searchable databases, then fetch the file blast.sdbs, and add the name of the local search set so that it appears in the menu.


[ Previous | Top | Next ]

You can set the parameters listed below from the command line. For more information, see "Using Program Parameters" in Chapter 3, Using Programs in the User's Guide.

Following some of the optional parameters described below is a letter or short expression in parentheses. These are the names of the corresponding parameters at the bottom of your BLAST output.

This parameter, for which there is a prompt if you don't set it on the command line, lets you influence the number of hits in your output having scores that would be expected to have occurred by chance alone. There is nothing to prevent many biologically significant but statistically insignificant segment pairs from being screened out, so you may sometimes want to increase this parameter in order to have an opportunity to see them.


By default, the BLAST output list file will contain 500 sequences (or fragments thereof, depending upon the state of -FRAgments), even if more than 500 sequences had scores above the cutoff score. The list is sorted in order of increasing probability, that is, with the most significant sequences first. Use -LIStsize to change the number of sequences in your output to any value between 0 (for blastall's program defaults) and 1000.


tells the program to use 2 threads for the database search on a multiprocessor computer. Check with your system manager for the number of processors available at your site. Never set the number of processors greater than what you have available.


When searching a nucleotide sequence database with a nucleotide query sequence, this specifies that tblastx should be run instead of blastn. tblastx translates the query and every sequence in the database and examines all pairwise combinations to find similarities at the amino acid level.

The search set menu can scroll off your screen if it contains all of the searchable databases supported locally on your computer. The next two parameters can reduce the size of that menu.

confines the menu to search sets containing nucleotide sequences.


confines the menu to search sets containing protein sequences.


sets the size of the short regions of similarity between sequences that BLAST initially searches for. If -WORdsize=0, BLAST uses the default values: 11 for blastn and 3 for the other programs. Smaller word sizes result in a more sensitive search at the expense of a longer search time.


sets the nucleotide match reward to 1 (blastn only).


sets the nucleotide mismatch penalty to -3 (blastn only).


sets the amino acid substitution matrix to use. BLAST normally uses the BLOSUM62 amino acid substitution matrix from Henikoff and Henikoff for protein sequence comparisons (including all cases where nucleotide database or query sequences are translated before comparison). Other valid options are BLOSUM45, BLOSUM80, PAM30, and PAM70.


sets the penalty for adding a gap to the alignment. See the RESTRICTIONS topic for more information about setting the gap opening penalty.


sets the penalty for lengthening an existing gap in the alignment. See the RESTRICTIONS topic for more information about setting the gap extension penalty.


sets the threshold for extending hits using the two-hit method. Words with scores at least this high can be extended as ungapped alignments.


suppresses filtering out low-complexity regions from query sequences.

By default, the SEG filter (Wootton and Federhen, Computers Chem. 17(2); 149-163 (1993)) masks low-complexity sequences in protein sequences, and the DUST filter (Hancock and Armstrong, Computer Applications in the Biosciences 10; 67-70 (1994)) performs the same function for nucleic acid sequences. Masked regions are excluded from the search. These regions are replaced with X's (in protein sequences) or N's (in nucleic acid sequences) in the output to let you identify the regions that were excluded.


sets a genetic code to use for the translation of the query sequence. BLAST uses the standard ("universal") genetic code unless you specify the number of one of the alternative codes listed under the topic ALTERNATIVE GENETIC CODES.


sets a genetic code to use for the translation of the database sequences. If you are searching for proteins from a system that doesn't use the standard ("universal") genetic code, you can select a more appropriate code from those listed under the topic ALTERNATIVE GENETIC CODES. Note that most of the genes in the nucleotide databases will be translated incorrectly if you select a nonstandard genetic code.


sets the effective database size. A value of 0 selects the program default.


suppresses the appearance of begin and end ranges on each output list file entry based on the alignment between the entry and the query sequence.


By default, BLAST displays the alignments of HSPs from the best 250 sequences in the list. Use -ALIgnments to change the number of sequences for which alignments are shown in your output to any value between 0 and 1000.


sets the alignment view type. Acceptable values are 0 through 8, which correspond to the following:

0 = pairwise (the default);
1 = showing identities as dots;
2 = showing insertions;
3 = showing identies as dots and gapping for insertions;
4 = gapping for insertions;
5 = with endgaps and showing insertions
6 = with endgaps flat master-slave and gapping for insertions
7 = XML output
8 = tab-delimited summary table
The specification of the XML output is available from NCBI at:

Here are descriptions of the columns in the tab-delimited format:

1 = Query sequence name
2 = Database sequence name
3 = Percent of positions that are identical
4 = Alignment length
5 = Number of mismatches (alignment length - identities - gapped positions)
6 = Number of gaps of any length
7 = Start of alignment for query sequence
8 = End of alignment for query sequence
9 = Start of alignment for database sequence
10 = End of alignment for database sequence
11 = Expectation
12 = Score (bits)

performs non-gapped alignments. By default gapped alignments are performed except when using tblastx, where gapped alignments are not available.

-XDRopoff=0 [X2]

sets the X2 dropoff value for gapped alignments (in bits). Gapped alignments are extended until the score drops below this value. This limits the (computationally expensive) extension of hits. Use -XDRopoff=0 for default behavior.


causes BLAST to use Miller's greedy algorithm to align sequences after performing the initial ungapped extension. You can only -MEGAblast when the type of the query sequence and database are both nucleotide.

This algorithm is optimized for aligning sequences that differ slightly as a result of sequencing or other similar errors and it can be considerably (up to 10 times) faster than BLASTN. As such, it is particularly useful for comparing large sequences. The minimum wordsize that is allowed with Megablast is larger than the standard default, which can reduce the sensitivity for relatively short sequences. See Z. Zheng et al. Journal of Computational Biology 7: 203-214 (2000) for more information regarding the greedy sequence alignment algorithm.

For the most part, any parameters that can be used with BLASTN can be used with -MEGAblast. However, unlike BLASTN, options that affect gapped extensions (e.g. XDRropoff) are ignored.


This parameter allows a user to search a nucleotide database with a protein query sequence using a PSSM from a previous PSIBLAST search as the scoring matrix. This kind of search is called PSI-TBLASTN. The PSSM is stored in a binary checkpoint file created by the program PSI-BLAST. Please see the PSIBLAST chapter in the Program Manual for an explanation of PSSM-based searching.

You must use the exact same query sequence that was used when the checkpoint file was saved.


masks lowercase characters in the query sequence by replacing them with the letter X during the search. Masked residues are ignored when calculating scores. This is one of the few cases in the Wisconsin Package where the uppercase and lowercase characters in input sequences can produce different results.


Sets the maximum distance allowed for two non-overlapping sequence segments on the same diagonal, when looking for matches between the query and a database sequence.


sets the maximum number of hits from a given region of the query sequence. Only the highest scoring hits from the region are kept. With -BESthits=0, the maximum number is set internally. This parameter can be used to counter the tendency of highly abundant, conserved regions to be so prevalent in the output that the detection of other domains would be precluded.


uses HTML format for output. This parameter has no effect if you use -VIEW=7 (XML output) or -VIEW=8 (tab-delimited output).


produces output in unmodified BLAST2 format.


The GCG Wisconsin Package implementation of BLAST is what is known as a "wrapper" program. After collecting your input parameters, the wrapper calls the locally-built implementation of BLAST from NCBI called blastall. If you are familiar with the interface to the blastall program as it was originally written, you can pass parameters to it directly using this parameter. Please call us if there are additional parameters you want to use with BLAST that you would like to look more like GCG parameters.


submits the program to the batch queue for processing after prompting you for all required user inputs. Any information that would normally appear on the screen while the program is running is written into a log file. Whether that log file is deleted, printed, or saved to your current directory depends on how your system manager has set up the command that submits this program to the batch queue. All output files are written to your current directory, unless you direct the output to another directory when you specify the output file.


lists valid databases then exits without searching.

The release notes for BLAST 2.0 can be found at

Printed: January 9, 2002 13:45 (1162)

[ Program Manual | User's Guide | Data Files | Databases ]

Technical Support:

Copyright (c) 1982-2002 Accelrys Inc. A subsidiary of Pharmacopeia, Inc. All rights reserved.

Licenses and Trademarks Wisconsin Package is a trademark and GCG and the GCG logo are registered trademarks of Accelrys Inc.

All other product names mentioned in this documentation may be trademarks, and if so, are trademarks or registered trademarks of their respective holders and are used in this documentation for identification purposes only.

Accelrys Inc.