Searches

Four search types are supported: phmmer, hmmsearch, hmmscan and jackhmmer. See HMMER algorithms for more information.

There are many different ways that a search on the website can be modified. Below is a list of the different accepted inputs and the parameters that can be modified. Also included are the parameter names that are required when using the API. This section is meant to be a guide to using the website, but further information can be found in the extensive HMMER guide. The parameter names used on the site are typically the same as the command line parameters, with the exception of the input data parameters. Each section is followed by a summary table that can be used as a quick reference.

Search query

phmmer, hmmscan and jackhmmer searches take a single protein amino acid sequence as the input, controlled by the seq parameter. The website accepts either FASTA format or an amino acid sequence. Alternatively, a sequence can be specified by accession or identifier. When using the website, suggestions will be offered as the name is typed.

Parameter name seq acc
Description Sets the query sequence
Algorithm(s) phmmer, hmmscan, jackhmmer
Accepted values Protein sequence (FASTA) Accession or identifier from one of the supported databases
Default None None
Required Yes (seq or acc)

hmmsearch and jackhmmer searches can take either a multiple protein sequence alignment as an input or a profile HMM. The alignment formats currently accepted are:

  • Aligned FASTA
  • Clustal (and Clustal-like)
  • PSI-BLAST
  • PHYLIP
  • Selex
  • GCG/MSF
  • STOCKHOLM format
  • UC Santa Cruz A2M (alignment to model)

The algorithms hmmsearch and jackhmmer also permit searches to be initiated with a profile HMM. This can be entered as text via the website, or via the seq or file parameters when using the API. Alternatively, it is also possible to retrieve HMMs from one of the supported HMM databases using the accession/identifier look up (in a similar manner to the sequence look up described earlier). To restrict the look up to one particular HMM database, append “@” followed by the database name (all lower case) e.g. CBS@pfam.

Query examples

For each of the search algorithms, examples sequences/alignments are provided (click on the ‘example’ button). These examples have been chosen to show a result set that demonstrates the various features available on the results pages.

Default search parameters

The searches on the website, when used in the simple mode, hide most of the search parameters and default values are used. Below is a list of the parameters and values used in the default search for each algorithm:

phmmer

Sequence database UniProt reference proteomes
Significance threshold (E-value) 0.01 for sequence matches; 0.03 for hit matches
Reporting threshold (E-value) 1 for both sequences and hits
Gap penalties open: 0.02; extend: 0.4; scoring matrix: BLOSUM62
Filter Bias composition filtering on
Pfam search Enabled, with gathering thresholds applied

hmmscan

HMM database Pfam
Significance threshold The Pfam gathering thresholds are used to determine hit significance
Filter Bias composition filtering on

hmmsearch

Sequence database UniProt reference proteomes
Significance threshold (E-value) 0.01 for sequence matches, 0.03 for hit matches
Reporting threshold (E-value) 1 for both sequences and hits
Filter Bias composition filtering on

jackhmmer

Sequence database UniProt reference proteomes
Significance threshold (E-value) 0.01 for sequence matches; 0.03 for hit matches
Reporting threshold (E-value) 1 for both sequences and hits
Gap penalties open: 0.02; extend: 0.4; scoring matrix: BLOSUM62
Filter Bias composition filtering on

Databases

Sequence databases

The sequence database field changes which target sequence database is searched. The default is UniProt references proteomes. This is one of the few parameters that is required by phmmer, hmmsearch or jackhmmer.

Parameter name seqdb
Description Sets the target sequence database
Algorithm phmmer, hmmsearch, jackhmmer
Accepted values uniprotrefprot, uniprotkb, swissprot, pdb, rp15, rp35, rp55, rp75, ensemblgenomes, ensembl, qfo
Default uniprotrefprot (see below)
Required Yes

HMM databases

This field indicates which profile HMM database the query should be searched against.

Parameter name hmmdb
Description Sets the target HMM database
Algorithm hmmscan
Accepted values gene3d, pfam, tigrfam, superfamily, pirsf
Default pfam
Required Yes

Thresholds

All four algorithms have the ability to set two different categories of cut-offs: significance and reporting thresholds. These cut-offs can be defined either as E-values (the default option) or bit scores. When setting either category of threshold, there are two values for each of the threshold categories: sequence and hit. A query can match a target in multiple places, defined as a hit (or domain) score. The sum of all hits on the sequence is the sequence score.

For example, trying to match repeating motifs can often be difficult, due to sequence variation in the repeating sequence motif. However, it can be possible to capture all examples of the motif, by relaxing the hit parameter while maintaining a stringent sequence parameter. This means that multiple matches, even if they are not strong matches, can be detected, but the sum of these matches must be sufficient to achieve the sequence score, there by limiting the rate of false positives.

Significance thresholds

Significance (or inclusion) thresholds are stricter than reporting thresholds and take precedence over them. These determine whether a sequence/hit is significant or not.

Significance E-values

Sequence and hit significance E-value thresholds will set matches with E-values less than or equal to the cut-off E-value as being significant (defaults below). If using the API, the incE and incdomE parameters are used to set the sequence and hit E-value thresholds respectively. In the absence of any threshold parameters the server will default to using E-value thresholds with the defaults.

Alternatively, the sequence and hit significance thresholds can be specified as bit scores. Any sequence or hit scoring greater than or equal to that given threshold will be considered a significant hit. By default, the form on the website is filled with typical values (defaults below). If using the API, the incT and incdomT parameters are used to set the sequence and hit bit thresholds respectively. This threshold is not used by default. If only one of these two parameters is set, then the unassigned parameter is set to the other assigned parameter value.

Parameter name incE incdomE
Description Sequence E-value threshold Hit E-value threshold
Algorithm phmmer, hmmscan, hmmsearch, jackhmmer
Accepted values 0 < x ≤ 10 10 < x ≤ 10
Default 0.01 or set to sequence threshold, if present 0.03 or set to hit threshold, if present
Required No No

Significance bit scores

Alternatively, the sequence and hit significance thresholds can be specified as bit scores. Any sequence or hit scoring greater than or equal to that given threshold will be considered a significant hit. By default, the form on the website is filled with typical values (defaults below). If using the API, the incT and incdomT parameters are used to set the sequence and hit bit thresholds respectively. This threshold is not used by default. If only one of these two parameters is set, then the unassigned parameter is set to the other assigned parameter value.

Parameter name incT incdomT
Description Sequence bit score threshold Hit bit score threshold
Algorithm phmmer, hmmscan, hmmsearch, jackhmmer
Accepted values x > 0 x > 0
Default 25.0 22.0
Required No No

Reporting thresholds

The reporting thresholds controls how many matches that fall below the significance threshold are still shown in the results (i.e. reported). As every entity in the target database is compared to the query, if all matches were reported, then potentially vast outputs would be generated. However, it can often be useful to view border-line matches as they may reveal more distant potential informative similarities to the model. As with the significance thresholds, there is a value for both the sequence and the hit, which again can be defined as either an E-value or a bit score. Such reported matches are indicated by a yellow background in the results table produced in the website.

Reporting E-values

Parameter name E domE
Description Sequence E-value threshold (reporting) Hit E-value threshold (reporting)
Algorithm phmmer, hmmscan, hmmsearch, jackhmmer
Accepted values 0 < x ≤ 10 10 < x ≤ 10
Default 1 or set to sequence threshold, if present 1 or set to hit threshold, if present
Required No No

Reporting bit scores

The sequence and hit reporting thresholds can also be specified as bit scores. Any sequence or hit scoring greater than or equal to that given threshold will be reported (defaults below). If using the API, the T and domT parameters are used to set the sequence and hit bit thresholds respectively. If significance thresholds are set, yet either or both reporting thresholds are undefined, these default form values will be set server side.

Parameter name T domT
Description Sequence E-value threshold (reporting) Hit E-value threshold (reporting)
Algorithm phmmer, hmmscan, hmmsearch, jackhmmer
Accepted values x > 0 x > 0
Default 7.0 5.0
Required No No

Gathering thresholds

Specific to hmmscan, the gathering threshold indicates to HMMER to use the sequence and hit thresholds defined in the HMM file to be searched. In the cases of Pfam and TIGRFAMs these are set conservatively to ensure that there are no known false positives. Thus, if a query sequence scores with a bit score greater than or equal to the gathering thresholds, then that match can be treated with high confidence. This threshold is the default setting for hmmscan. If you are using the API, you can use the cut_ga parameter to signify that the gathering threshold should be used.

Gene3D and Superfamily thresholds

Both of these HMM databases apply sophisticated post-processing steps on the HMMER results to make the domain assignments and disentangle overlapping matches. Each database uses an internal E-value cut-off of 0.0001 for a domain match and does not employ the use of HMM specific bit score thresholds. Thus, cut-off manipulation has been disabled for these databases, thereby faithfully replicating the results of these HMM databases.

Advanced search options

Taxonomy Restrictions

Pre-defined Taxonomic Tree

You can select different levels of a given taxonomic tree. All species within the selected levels will be included in your search.

Customisation of results

The result table may be customised to display different columns and/or to restrict the number of rows in the table to a manageable number. This can be performed before or after the search, with the customisation stored in a cookie so that you will not have to keep re-configuring the table after each search.

Filters

Bias composition

Turning off the bias composition filter can increases sensitivity, but at a high cost in speed, especially if the query has biased residue composition (such as a repetitive sequence region, or a membrane protein with large regions of hydrophobicity). Without the bias filter, too many sequences may pass the filter with biased queries, leading to slower than expected performance, hence it is switched on by default. This feature can be disabled using the nobias parameter.

Parameter name nobias
Description Turns off the bias composition filtering
Algorithms phmmer, hmmscan, hmmsearch, jackhmmer
Accepted Values 1
Required No

Gap penalties

These are specific to phmmer and jackhmmer (initiated with a single sequence).

Open

The open parameter (called popen in HMMER) sets the probability for opening a gap in an alignment between target sequence against the model (or query sequence). The default value is 0.02, but can be set any value from 0 (no gaps) to less than 0.5 (more likely to extend the gap).

Extend

The extend parameter (called pextend in HMMER) sets the probability for extending the gap for a target sequence against the model or query sequence. The default value is 0.4, but can be set anywhere from 0 (less likely to extend) to less than 1 (more likely to extend the gap).

Scoring Matrix

When using phmmer, the query is a single sequence so the residue alignment probabilities are calculated from a substitution matrix. Substitution matrices provide scores that indicate the likelihood of two aligned amino acids appearing due to conservation rather than by chance. There are five different matrices available for selection: BLOSUM45, BLOSUM62 (default), BLOSUM90, PAM30 and PAM70. These BLOSUM matrices are based on observed alignments between amino acids in the BLOCKS database, where as the PAM matrices have been extrapolated from comparisons of closely related proteins. The different matrices alter the stringency of the alignment e.g. PAM90 can be used to find more distantly related sequences than PAM70, as PAM70 is more stringent; BLOSUM62 can be used to find more closely related sequence than using BLOSUM45, as BLOSUM45 is less stringent.

This is required for phmmer and jackhmmer and default values will be used if no value is set.

Parameter name popen pextend mx
Description Gap open penalty Gap extend penalty Substitution matrix
Algorithm(s) phmmer, jackhmmer
Accepted values 0 ≤ x < 0.5 0 ≤ x < 1 BLOSUM45, BLOSUM62, BLOSUM90, PAM30, PAM70
Default 0.02 0.4 BLOSUM62
Required No

Batch searches

It is also possible to search multiple protein sequences in ‘offline’ batch mode. With both phmmer and hmmscan, files containing sequences in FASTA format can be uploaded via the “Upload a file” link. These sequences will then be searched, in turn, against the specified databases. There is a limit of 500 sequences per batch request. This is only to prevent overload of the servers: multiple batch requests are permitted. Once the job is submitted, a different results page will be returned, showing a table with each row in that table representing a sequence in your file. This table periodically updates, indicating the progress of your batch job. As results appear in the table, you can view the details. If you have many sequences, you can also request that an e-mail be sent when the batch job has completed. It is also possible to use hmmsearch in batch mode, again with a single multiple alignment or profile HMM.

The jackhmmer batch system operates in a slightly different manner. Under the advance settings you can select the number of iterations to be performed and the batch mode will automaticaly run through each iteration (or until convergence), taking the results and using all the sequences scoring above the significance threholds to generate the input multiple sequnece alignment for the next round. Only one sequence, multiple sequence aligment or profile HMM can be submitted at a time.

The batch system also works via the API, except the seq parameter is substituted for the file parameter; the other parameters remain the same. Requesting an e-mail notification can be set using the email parameter.

Glossary

Bit score
A bit score in HMMER is the log of the ratio of the sequence’s probability according to the profile (the homology hypothesis) to the null model probability (the non-homology hypothesis).
E-value
An E-value (expectation value) is the number of hits that would be expected to have a score equal to or better than this by chance alone. A good E-value is much less than 1, for example, an E-value of 0.01 would mean that on average about 1 false positive would be expected in every 100 searches with different query sequences. An E-value around 1 is what we expect just by chance. E-values are widely used as all you need to decide on the significance of a match is the E-value, but note that they vary according to the size of the target database.
Gathering threshold
Also called the gathering cut-off, the gathering threshold is actually comprised of two bit scores, a sequence cut-off and a domain cut-off, used to define the significance of a sequence and a hit respectively. These are defined in the profile HMM and set both significance and reporting thresholds so that no insignificant hits are reported.
Null model
The “null model” calculates the probability that the target sequence is not homologous to the query profile and is a one-state HMM configured to generate “random” sequences of the same mean length L as the target sequence, with each residue drawn from a background frequency distribution (a standard i.i.d. model: residues are treated as independent and identically distributed). This background frequency is based on the mean residue frequencies in Swiss-Prot 50.8 (October 2006).
Profile HMM
Profile hidden Markov Models (HMMs) are a way of turning a multiple sequence alignment into a position-specific scoring system, which is suitable for searching databases for remotely homologous sequences.
STOCKHOLM format
STOCKHOLM format is a multiple sequence alignment format supported by HMMER.