Features and Annotations HOWTO



Authors

Brian Osborne briano at bioteam.net

Steve Chervitz sac at bioperl.org

This document is copyright Brian Osborne. It can be copied and distributed under the terms of the Perl Artistic License.

Abstract

This is a HOWTO that explains how to use the SeqFeature and Annotation objects of Bioperl.

Introduction

There’s no more central notion in bioinformatics than the idea that portions of protein or nucleotide sequence have specific characteristics (or features). A given stretch of DNA may have been found to be essential for the proper transcriptional regulation of a gene, or a particular amino acid sequence may bind a particular ion, for example. This simple idea turns out to be a bit more complicated in the bioinformatics world where there’s a need to represent the actual data in all its varied forms. The promoter region may not be precisely defined down to the base pair, a transcribed region may be divided into discontinuous exons, a gene may have different numbered positions on different maps, a sequence may have a sub-sequence which itself possesses some characteristic, an experimental observation may be associated with a literature reference, and so on.

This HOWTO describes aspects of Bioperl’s approach. The problem is how to create software that accepts, analyzes, and displays any and all of this sequence annotation with the required attention to detail yet remains flexible and easy to use. The general names for the modules or objects that serve these purposes in Bioperl are Bio::SeqFeature and Bio::Annotation.

The HOWTO will discuss these objects and the differences between them. There’s also discussion of how to get useful data from these objects and the basics of how to create your own sequences using the objects.

The Basics

Some BioPerl neophytes may also be new to object-oriented programming (OOP) and this notion of an object. OOP is not the subject of this HOWTO but there should be some discussion of how objects are used in BioPerl. In the BioPerl world parsing a GenBank file doesn’t give you data, it gives you an object and you can ask the object, a kind of variable, for data. While annotating you don’t create a file or database entry directly. You might create a sequence object and an annotation object, then put these two together to create an annotated sequence object. You could then tell this object to make a version of itself as a file, or pass this object to a database object in order to enter some data into the database. This is a very flexible and logical way to design a complex piece of software like BioPerl, since each part of the system can be created and evaluated separately.

A central idea in OOP is inheritance, which means that a child object can derive some of its capabilities or functionality from a parent object. The OOP approach also allows new modules to modify or add functionality, distinct from the parent. Practically speaking this means that there’s not one definitive SeqFeature or Annotation object but many, each a variation on a theme. The details of the these varieties will be discussed in other sections, but for now we could use some broad definitions that apply to all the variations.

Features

A Bio::SeqFeature object is designed to be associated with a sequence, and can have a location on that sequence - it’s a way of describing the characteristics of a specific part of a sequence. SeqFeature objects can also have features themselves, which you could call sub-features but which, in fact, are complete SeqFeature objects. SeqFeature objects can also have one or more Annotations associated with them.

Annotations

An Bio::Annotation object is also associated with a sequence as you’d expect but it does not have a location on the sequence, it’s associated with an entire sequence. This is one of the important differences between a SeqFeature and an Annotation. Annotations also can’t have SeqFeatures, which makes sense since SeqFeature objects must have locations. The relative simplicity of the Annotation has made it amenable to the creation of a useful set of Annotation objects, each devoted to a particular kind of fact or observation.

Locations were discussed, above. Describing locations can be complicated in certain situations, say when some feature is located on different sequences with varying degrees of precision. One location could also be shared between disparate objects, such as two different kinds of SeqFeatures. You may also want to describe a feature with many locations, like a repeated sequence motif in a protein. Because of these sorts of complexities and because one may want to create different types of locations the BioPerl authors elected to keep location functionality inside dedicated Bio::Location objects.

SeqFeatures and Annotations will make the most sense if you’re already somewhat familiar with BioPerl and its objects. The reader is referred to the Beginners HOWTO, SeqIO HOWTO, and the SearchIO HOWTO for more information on these topics. Here’s a bit of code, to summarize:

# BAB55667.gb is a Genbank file, and Bioperl knows that it
# is a Genbank file because of the '.gb' file suffix
use Bio::SeqIO;
my $seqio_object = Bio::SeqIO->new(-file => "BAB55667.gb" );
my $seq_object = $seqio_object->next_seq;

Now that we have a sequence object, $seq_object, we can examine its features and annotations.

Getting the Features

The focus of this HOWTO is mostly on Genbank format but bear in mind that all of the code shown here will also work on other formats containing features and annotations (EMBL, Swissprot, BSML, Chado XML, GAME, KEGG, Locuslink, Entrez Gene, TIGR XML). When the entry comes from Genbank it’s easy to see where most of the features are, they’re in the FEATURES table section, something like this:

 FEATURES            Location/Qualifiers
     source          1..1846
                     /organism="Homo sapiens"
                     /db_xref="taxon:9606"
                     /chromosome="X"
                     /map="Xp11.4"
     gene            1..1846
                     /gene="NDP"
                     /note="ND"
                     /db_xref="LocusID:4693"
                     /db_xref="MIM:310600"
     CDS             409..810
                     /gene="NDP"
                     /note="Norrie disease (norrin)"
                     /codon_start=1
                     /product="Norrie disease protein"
                     /protein_id="NP_000257.1"
                     /db_xref="GI:4557789"
                     /db_xref="LocusID:4693"
                     /db_xref="MIM:310600"
                     /translation="MRKHVLAASFSMLSLLVIMGDTDSKTDSSFIMDSDPRRCMRHHY
                     VDSISHPLYKCSSKMVLLARCEGHCSQASRSEPLVSFSTVLKQPFRSSCHCCRPQTSK
                     LKALRLRCSGGMRLTATYRYILSCHCEECNS"

Features in Bioperl are accessed using their tags, either a primary tag or a plain tag. Examples of primary tags and tags in this Genbank entry are shown below. You can see that in this case the primary tag is a means to access the tags and it’s the tags that are directly associated with the data from the file.

Tag name Tag type Tag value
source primary tag  
CDS primary tag  
gene primary tag  
organism tag Homo sapiens
note tag ND
protein_id tag NP_000257.1
translation tag MRKHVL…HCEECNS
db_xref tag MIM:310600

Table 1. Tag examples

When a Genbank file like the one above is parsed the feature data is converted into objects, specifically Bio::SeqFeature::Generic objects. How many? In this case 3, one for each of the primary tags.

In other parts of the Bioperl documentation one finds discussions of the “SeqFeature object”, but there’s more than one kind of these, as we’ll see later, so what is this a reference to? More than likely it’s referring to this same Bio::SeqFeature::Generic object. Think of it as the default SeqFeature object. Now, should you care what kind of object is being made? For the most part no, you can write lots of useful and powerful Bioperl code without ever knowing these specific details.

By the way, how does one know what kind of object one has in hand? Try something like:

print ref($seq_object);
# results in "Bio::Seq::RichSeq"

The Bio::SeqFeature::Generic object uses tag/value pairs to store information, and the values are always returned as arrays. A simple way to access all the data in the features of a Seq object would look something like this:

for my $feat_object ($seq_object->get_SeqFeatures) {
    print "primary tag: ", $feat_object->primary_tag, "\n";

    for my $tag ($feat_object->get_all_tags) {
        print "  tag: ", $tag, "\n";
        for my $value ($feat_object->get_tag_values($tag)) {
            print "    value: ", $value, "\n";
        }
    }
}

This bit would print out something like:

 primary tag: source
   tag: chromosome
     value: X
   tag: db_xref
     value: taxon:9606
   tag: map
     value: Xp11.4
   tag: organism
     value: Homo sapiens
 primary tag: gene
   tag: gene
     value: NDP
   tag: note
     value: ND
 primary tag: CDS
   tag: codon_start
     value: 1
   tag: db_xref
     value: GI:4557789
     value: LocusID:4693
     value: MIM:310600
   tag: product
     value: Norrie disease protein
   tag: protein_id
     value: NP_000257.1
   tag: translation
     value: MRKHVLAASFSMLSLLVIMGDTDSKTDSSFIMDSDPRRCMRHHYVDSI
            SHPLYKCSSKMVLLARCEGHCSQASRSEPLVSFSTVLKQPFRSSCHCC
            RPQTSKLKALRLRCSGGMRLTATYRYILSCHCEECNS

So to retrieve specific values, like all the database identifiers, you could do:

for my $feat_object ($seq_object->get_SeqFeatures) {
    push @ids, $feat_object->get_tag_values("db_xref") 
       if ($feat_object->has_tag("db_xref"));
}

Important Make sure to include that if ($feat_object->has_tag("...")) part, otherwise you’ll get errors when the feature does not have the tag you’re requesting.

One last note on Genbank features. The Bioperl parsers for Genbank and EMBL are built to respect the specification for the feature tables agreed upon by Genbank, EMBL, and DDBJ (see the Feature Table Definition for the details). Check this page if you’re interested in a complete listing and description of all the Genbank, EMBL, and DDBJ feature tags.

Despite this specification some non-standard feature tags have crept into Genbank, like bond. When the Bioperl SeqIO Genbank parser encounters a non-standard feature like this it’s going to throw a fatal exception. The work-around is to use eval{} so your script doesn’t die, something like:

use Bio::SeqIO;

my $seq_object;
my $seqio_object = Bio::SeqIO->new(-file => $gb_file,
                                   -format => "genbank");

eval { $seq_object = $seqio_object->next_seq; };

# if there's an error
print "Problem in $gb_file. Bad feature perhaps?\n" if $@;

Getting Sequences

One commonly asked question is “How do I get the sequence of a SeqFeature?” The answer is “It depends on what you’re looking for.” If you’d like the sequence of the parent, the sequence object that the SeqFeature is associated with, then use entire_seq():

$seq_object = $feat_object->entire_seq;

This doesn’t return the parent’s sequence directly but rather a Bio::PrimarySeq object corresponding to the parent sequence. Now that you have this object you can call its seq() method to get the sequence string, or you could do this all in one step:

my $sequence_string = $feat_object->entire_seq->seq;

There are 2 other useful methods, seq() and spliced_seq(). Consider the following Genbank example:

 FEATURES             Location/Qualifiers
      source          1..177
                      /organism="Mus musculus"
                      /mol_type="genomic DNA"
                      /db_xref="taxon:10090"
      tRNA            join(103..111,121..157)
                      /gene="Phe-tRNA"

To get the sequence string from the start to the end of the tRNA feature use seq(). To get the spliced sequence string, accounting for the start and end locations of each sub-sequence, use spliced_seq(). Here are the methods and the corresponding example coordinates:

Method Coordinates
entire_seq() 1..177
seq() 103..157
spliced_seq() 103..111,121..157

Table 2. Sequence retrieval methods

It’s not unusual for a Genbank file to have multiple CDS or gene features (and recall that CDS and gene are common primary tags in Genbank format), each with a number of tags, like note, protein_id, or product. How can we get, say, the nucleotide sequences and gene names from all these CDS features? By putting all of this together we arrive at something like:

use Bio::SeqIO;

my $seqio_object = Bio::SeqIO->new(-file => $gb_file);
my $seq_object = $seqio_object->next_seq;

for my $feat_object ($seq_object->get_SeqFeatures) {
    if ($feat_object->primary_tag eq "CDS") {
        print $feat_object->spliced_seq->seq,"\n";
        # e.g. 'ATTATTTTCGCTCGCTTCTCGCGCTTTTTGAGATAAGGTCGCGT...'
        
        if ($feat_object->has_tag('gene')) {
            for my $val ($feat_object->get_tag_values('gene')) {
                print "gene: ",$val,"\n";
                # e.g. 'NDP', from a line like '/gene="NDP"'
            }
        }
    }
}

Compact Code

Many people wouldn’t write code in the rather deliberate style used above. The following is more compact code that gets all the features with a primary tag of CDS, starting with a Genbank file:

my @cds_features = grep { $_->primary_tag eq 'CDS' } 
 Bio::SeqIO->new(-file => $gb_file)->next_seq->get_SeqFeatures;

With this array of SeqFeatures you could do all sorts of useful things, such as find all the values for the gene tags and their corresponding spliced nucleotide sequences and store them in a hash:

my %gene_sequences = map {$_->get_tag_values('gene'), 
    $_->spliced_seq->seq } @cds_features;

Because you’re asking for a specific primary tag and tag, CDS and gene respectively, this code would only work when there are features that looked something like this:

     CDS             735..182
                     /gene="MG001
                     /codon_start=
                     /product="DNA polymerase III, subunit beta (dnaN)
                     /protein_id="AAC71217.1
                     /translation="MNNVIISNNKIKPHHSYFLIEAKEKEINFYANNEYFSVKCNLN
                     NIDILEQGSLIVKGKIFNDLINGIKEEIITIQEKDQTLLVKTKKTSINLNTINVNEF
                     RIRFNEKNDLSEFNQFKINYSLLVKGIKKIFHSVSNNREISSKFNGVNFNGSNGKEI
                     LEASDTYKLSVFEIKQETEPFDFILESNLLSFINSFNPEEDKSIVFYYRKDNKDSFS
                     EMLISMDNFMISYTSVNEKFPEVNYFFEFEPETKIVVQKNELKDALQRIQTLAQNER
                     FLCDMQINSSELKIRAIVNNIGNSLEEISCLKFEGYKLNISFNPSSLLDHIESFESN
                     INFDFQGNSKYFLITSKSEPELKQILVPSR

Location Objects

There’s quite a bit to this idea of location, so much that it probably deserves its own HOWTO. Another way of saying this is that if this topic interests you should take a closer look at the modules that are concerned with both Location and Range such as Bio::Range, Bio::RangeI, and Bio::LocationI. The Range object is the simpler of the two, it holds the start, end, and strand (1, -1) information for a sequence that is located on some other sequence, typically a larger one. The Range object can only describe exact locations.

The Location object is a Range object but it has additional capabilities designed to handle inexact or fuzzy locations, where the start and end of a particular sub-sequence themselves have start and end positions, or are not precisely defined.

Both these objects use methods like overlaps(), contains(), union() and intersection() that act on pairs of Ranges or Locations. The table below is meant to illustrate some of the modules’ descriptive capabilities.

Type Example
EXACT (5..100)
BEFORE (<5..100)
AFTER (>5..100)
WITHIN ((5.10)..100)
BETWEEN (99^100)
UNCERTAIN (99.?100)

Table 3. Location examples

One type that might not be self-explanatory is WITHIN. The example means starting somewhere between positions 5 and 10, inclusive, and ending at 100. BETWEEN is interesting - the example means between 99 and 100, exclusive. A biological example of such a location would be a cleavage site, between two bases or residues, but not including them.

The UNCERTAIN attribute means what it says, not known. This value is found occasionally in SwissProt features.

In their simplest form the Location and Range objects are used to get or set start and end positions, getting the positions could look like this:

       # polyA_signal    1811..1815 
       #                 /gene="NDP"
       my $start = $feat_object->location->start;
       my $end = $feat_object->location->end;

By now you’ve figured out that the location() method returns a Location object - this object has end() and start() methods.

Another way of describing a feature in Genbank involves multiple start and end positions. These could be called split locations, and a very common example is the join statement in the CDS feature found in Genbank entries (e.g. join(45..122,233..267)). This calls for a specialized object, Bio::Location::SplitLocationI, which is a container for Location objects:

for my $feature ($seqobj->top_SeqFeatures){

    if ( $feature->location->isa('Bio::Location::SplitLocationI')
         && $feature->primary_tag eq 'CDS' )  {
         
        for my $location ( $feature->location->sub_Location ) {
            print $location->start . ".." . $location->end . "\n";
        }

    }
}

The Species Object

Note Future use of Bio::Species beyond release 1.6 is deprecated.

Some data in a Genbank file is accessible both as a feature and through a specialized object. Taxonomic information on a sequence, below, can be accessed through a Species object as well as a value to the “organism” tag, and you’ll get more information from the Bio::Species object. The taxonomic information for sequence looks like this in GenBank format:

 SOURCE      human.
   ORGANISM  Homo sapiens
             Eukaryota; Metazoa; Chordata; Craniata; Vertebrata; Euteleostomi;
             Mammalia; Eutheria; Primates; Catarrhini; Hominidae; Homo.

To access this data you’ll need to get a Species object from the Sequence object, and then use its methods:

# legible and long
my $species_object = $seq_object->species;
my $species_string = $species_object->node_name;

# Perlish
my $species_string = $seq_object->species->node_name;

# either way, $species_string is "Homo sapiens"

# get all taxa from the ORGANISM section in an array
my @classification = $seq_object->species->classification;

# "sapiens Homo Hominidae Catarrhini Primates Eutheria Mammalia
# Euteleostomi Vertebrata Craniata Chordata Metazoa Eukaryota"

The reason that ORGANISM isn’t treated only as a plain tag is that there are a variety of things one would want to do with taxonomic information, so returning just an array wouldn’t suffice. See the documentation on for more information on its methods.

Getting the Annotations

There’s still quite a bit of data left in our Genbank files that’s not in SeqFeature objects, and much of it is parsed into Bio::Annotation objects. Annotations, if you recall, are those values that are assigned to a sequence that have no specific location on that sequence. In order to get access to these objects we will get an Bio::AnnotationCollection object, which is exactly what it sounds like:

my $io = Bio::SeqIO->new(-file => $file, -format => "genbank" );
my $seq_obj = $io->next_seq; my $anno_collection = $seq_obj->annotation;

Now we can access each Bio::Annotation in the Bio::AnnotationCollection object. The Annotation objects can be retrieved in arrays:

for my $key ( $anno_collection->get_all_annotation_keys ) {

    my @annotations = $anno_collection->get_Annotations($key);
      
    for my $value ( @annotations ) {
        print "tagname : ", $value->tagname, "\n";
        
        # $value is an Bio::Annotation, and also has an "as_text" method
        print "  annotation value: ", $value->display_text, "\n";
    }
}

It turns out the value of $key, above, and $value->tagname are the same. The code will print something like:

    tagname : comment
      annotation value: Comment: REVIEWED REFSEQ: This record has been curated by NCBI staff. The reference sequence was derived from X65882.1. Summary: NDP is the genetic locus identified as harboring mutations that result in Norrie disease.
    tagname : reference
      annotation value: Reference: The molecular biology of Norrie's disease
    tagname : date_changed
      annotation value: Value: 31-OCT-2000

If you only wanted a specific annotation, like COMMENT, you can use the tag name as an argument:

my @annotations = $anno_collection->get_Annotations('comment');

And if you’d simply like all of the Annotations, regardless of key, you can do this:

my @annotations = $anno_collection->get_Annotations();

The following is a table of some of the common Annotations, their keys in Bioperl, and what they’re derived from in Genbank files:

GenBank Text Key Object Type Note
COMMENT comment Comment  
SEGMENT segment SimpleValue e.g. “1 of 2”
ORIGIN origin SimpleValue e.g. “X Chromosome.”
REFERENCE reference Reference  
INV date_changed SimpleValue e.g. “08-JUL-1994”
KEYWORDS keyword SimpleValue  
ACCESSION secondary_accession SimpleValue 2nd of 2 accessions
DBSOURCE dblink DBLink Link to entry in a database

Table 4. GenBank Annotation

Some Annotation objects, like Reference, make use of a hash_tree() method, which returns a hash reference. This is a more thorough way to look at the actual values than the display_text() method used above. For example, display_text() for a Reference object is only going to return the title of the reference, whereas the keys of the hash from hash_tree() will be title, authors, location, medline, start, and end.

if ($value->tagname eq "Reference") {
    my $hash_ref = $value->hash_tree;
    for my $key (keys %{$hash_ref}) {
         print $key,": ",$hash_ref->{$key},"\n";
    }
}

Which yields:

 authors: Meitinger,T., Meindl,A., Bork,P., Rost,B., Sander,C., Haasemann,M. and Murken,J.
 location: Nat. Genet. 5 (4), 376-380 (1993) 
 medline: 94129616
 title: Molecular modelling of the Norrie disease protein predicts a cystine knot
 growth factor tertiary structure
 end: 1846
 start: 1

Other Annotation objects, like Bio::Annotation::SimpleValue, also have a hash_tree() method but the hash isn’t populated with data and display_text() will suffice.

The simplest bits of Genbank text, like KEYWORDS, end up in these Bio::Annotation::SimpleValue objects, the COMMENT ends up in a Bio::Annotation::Comment object, and references are transformed into Bio::Annotation::Reference objects. Some of these specialized objects will have specialized methods. Take the Bio::Annotation::Reference object, for example:

if ($value->tagname eq "reference") {
    print "author: ",$value->authors(), "\n";
}

There’s also title(), publisher(), medline(), editors(), database(), pubmed() and a number of other methods.

Directly From the Sequence Object

This is just a reminder that some of the annotation data in your sequence files can be accessed directly, without looking at SeqFeatures or Annotations.

For example, if the Sequence object in hand is a Bio::Seq::RichSeq object then here are some useful methods:

Method Returns
get_secondary_accessions array
keywords array
get_dates array
seq_version string
pid string
division string

Table 5. Bio::Seq::RichSeq methods

These objects are created automatically when you use to read from EMBL, GenBank,GAME, Chado XML, TIGR XML, Locuslink, BSML, KEGG, Entrez Gene, and SwissProt sequence files. However, it’s not guaranteed that each of these formats will supply data for all of the methods above. See the SeqIO HOWTO for more details on formats.

Other Sequence File Formats

It is worth mentioning other sequence file formats. The table below shows what sorts of objects, Annotation or SeqFeature, you’ll get when you parse other sequence formats using .

Format SeqIO Name SeqFeature Annotation
Genbank genbank yes yes
EMBL embl yes yes
GAME game yes no
Chado XML chadoxml yes yes
TIGR XML tigr yes yes
Locuslink locuslink no yes
BSML bsml yes yes
KEGG kegg yes yes
SwissProt swiss yes yes
Entrez Gene entrezgene no yes

Table 6. Formats, SeqFeature’s, and Annotations

How does one find out what data is in which object in these formats? In general the individual module documentation is not going to provide all the answers, you’ll need to do some investigation yourself. Let’s use an approach we used earlier to dissect a Locuslink entry in a file, 148.ll. Here’s the file:

 LOCUSID: 148
 LOCUS_CONFIRMED: yes
 LOCUS_TYPE: gene with protein product, function known or inferred 
 ORGANISM: Homo sapiens
 STATUS: REVIEWED 
 NM: NM_000680|4501960|na
 NP: NP_000671|4501961
 PROT: AAA93114|409029
 ACCNUM: M11313|177869|na|na|na
 TYPE: p
 PROT: P35348|1168246
 OFFICIAL_SYMBOL: ADRA1A
 OFFICIAL_GENE_NAME: adrenergic, alpha-1A-, receptor
 ALIAS_SYMBOL: ADRA1C
 SUMMARY: Summary: Alpha-1-ARs are members of the GPCR superfamily.
 CHR: 8
 STS: SGC35557|8|8124|na|seq_map|epcr 
 COMP: 10090|Adra1a|14|14  cM|11549|8|ADRA1A|ncbi_mgd
 ALIAS_PROT: adrenergic, alpha-1C-, receptor
 BUTTON: unigene.gif
 LINK: http://www.ncbi.nlm.nih.gov/UniGene/clust.cgi?ORG=Hs&CID=52931
 UNIGENE: Hs.52931
 OMIM: 104221
 MAP: 8p21-p11.2|RefSeq|C|
 MAPLINK: default_human_gene|ADRA1A
 GO: cellular component|integral to plasma membrane|P|[GO:0005887|Proteome|8396931](GO:0005887%7CProteome%7C8396931)

First collect all the annotations:

use Bio::SeqIO;

my @annotations = Bio::SeqIO->new(-file => "148.ll", 
   -format => "locuslink")->next_seq->annotation->get_Annotations;

And from this array of Annotations let’s extract a hash containing the as_text strings as keys and the concatenated tagnames and object types as values:

my %tagname_type = map {$_->as_text,($_->tagname . " " . ref($_)) } 
  @annotations;

The contents of the %tagname_type hash can be represented in table form, below.

as_text tagname ref    
Direct database link to AAA93114 in database GenBank dblink Bio::Annotation::DBLink    
Value: http://www.ncbi.nlm.nih.gov/UniGene/clust.cgi?ORG=Hs&CID=52931 URL Bio::Annotation::SimpleValue    
Value: 8 CHR Bio::Annotation::SimpleValue    
Direct database link to NP_000671 in database RefSeq dblink Bio::Annotation::DBLink    
Direct database link to SGC35558 in database STS dblink Bio::Annotation::DBLink    
Comment: Summary: Alpha-1-ARs are members of the GPCR superfamily comment Bio::Annotation::Comment    
Value: adrenergic, alpha-1A-, receptor OFFICIAL_GENE_NAME Bio::Annotation::SimpleValue    
Value: ADRA1C ALIAS_SYMBOL Bio::Annotation::SimpleValue    
Value: adrenergic, alpha -1A-, receptor ALIAS_PROT Bio::Annotation::SimpleValue    
Direct database link to NM_000680 in database RefSeq dblink Bio::Annotation::DBLink    
Value: ADRA1A OFFICIAL_SYMBOL Bio::Annotation::SimpleValue    
Direct database link to SGC35557 in database STS dblink Bio::Annotation::DBLink    
Value: 8p21-p11.2 MAP Bio::Annotation::SimpleValue    
Direct database link to 104221 in database MIM dblink Bio::Annotation::DBLink    
Direct database link to D8S2033 in database STS dblink Bio::Annotation::DBLink    
Direct database link to none in database GenBank dblink Bio::Annotation::DBLink    
cellular component integral to plasma membrane cellular component Bio::Annotation::OntologyTerm
Direct database link to Hs.52931 in database UniGene dblink Bio::Annotation::DBLink    
Direct database link to M11313 in database GenBank dblink Bio::Annotation::DBLink    
Direct database link to P35348 in database GenBank dblink Bio::Annotation::DBLink    

Table 7. LocusLink Annotations

The output from the script shows that Locuslink Annotations come in a variety of types, including DBLink, OntologyTerm, Comment, and SimpleValue. In order to extract the exact value you want, as opposed to the one returned by the as_text method, you’ll need to find the desired method in the documentation for the Annotation in question.

If you were only interested in a certain type of Annotation you could retrieve it efficently with something like this:

@ontology_terms = map { $_->isa("Bio::Ontology::TermI"); } 
  $seq_object->get_Annotations();

To completely parse these sequence formats you may also need to use methods that don’t have anything to do with Features or Annotations per se. For example, the display_id method returns the LOCUS name of a Genbank entry or the ID from a SwissProt file. The desc() method will return the DEFINITION line of a Genbank file or the DE field in a SwissProt file. Again, this is a situation where you may have to examine a module, probably a SeqIO module, to find out more of the details.

Building Your Own Sequences

We’ve taken a look at getting data from SeqFeature and Annotation objects, but what about creating these objects when you already have the data? The Bio::SeqFeature::Generic object is probably the best SeqFeature object for this purpose, in part because of its flexibility. Let’s assume we have a sequence that has an interesting sub-sequence, going from position 10 to 22 on the + or 1 or sense strand.

use Bio::SeqFeature::Generic;

# create the feature with some data, evidence and a note
my $feat = new Bio::SeqFeature::Generic(-start  => 10,
                                        -end    => 22,
                                        -strand => 1,
                                        -primary_tag => 'TATA_signal',
                                        -tag => {evidence => 'predicted',
                                                 note     => 'TATA box' } );

The SeqFeature::Generic object offers the user a tag system for addition of data that’s not explicitly accounted for by its methods, that’s what the -tag is for, above. Since the value passed to -tag could be any kind of scalar, like a reference, it’s clear that this approach should be able to handle just about any sort of data.

You can build on the Feature as well. Here we’ll add some Annotations to the newly-created Feature:

$feat->add_tag_value("match1","PF000123 e-7.2");
$feat->add_tag_value("match2","PF002534 e-3.1");

my @tags = $feat->get_all_tags;

for my $tag (@tags) {
    for my $val ( $feat->get_tag_values($tag) ) {
       print $tag,":",$val,"\n";
    }
}

This prints out:

  evidence:predicted
  match1:PF000123 e-7.2
  match2:PF002534 e-3.1
  note:TATA box

Note If you need to add a tag that don’t have any value when printed (like /pseudo, /trans_splicing, or /environmental_sample), you can use the special value _no_value to make BioPerl print the tag without any associated value:

$feat->add_tag_value("pseudo","_no_value");

Once the feature and its annotations are created it can be associated with a sequence:

use Bio::Seq;

# create a simple Sequence object
my $seq_obj = Bio::Seq->new(-seq => "attcccccttataaaattttttttttgaggggtggg",
                            -display_id => "BIO52" );

# then add the feature we've created to the sequence
$seq_obj->add_SeqFeature($feat);

The add_SeqFeature() method will also accept an array of SeqFeature objects.

What if you wanted to add an Annotation to a sequence? You’ll create the Annotation object, add data to it, create an object, add the Annotation to the AnnotationCollection along with a tag, and then add the AnnotationCollection to the sequence object:

use Bio::Annotation::Collection;
use Bio::Annotation::Comment;

my $comment = Bio::Annotation::Comment->new;
$comment->text("This looks like a good TATA box");
my $coll = new Bio::Annotation::Collection;
$coll->add_Annotation('comment',$comment);
$seq_obj->annotation($coll);

Now let’s examine what we’ve created by writing the contents of $seq_obj to a Genbank file called test.gb. We should see a sequence, an Annotation associated with the sequence, a sequence Feature, and tag/value pairs associated with the Feature:

use Bio::SeqIO;
my $io = Bio::SeqIO->new(-format => "genbank",
                         -file => ">test.gb" );
                            
$io->write_seq($seq_obj);

Voila! test.gb now reads:

 LOCUS       BIO52                    36 bp    dna     linear   UNK
 ACCESSION   unknown
 COMMENT     This looks like a good TATA box
 FEATURES             Location/Qualifiers
      TATA_signal     10..22
                      /match2="PF002534 e-3.1"
                      /match1="PF000123 e-7.2"
                      /evidence=predicted
                      /note="TATA box"
                      /pseudo
 ORIGIN
         1 attccccctt ataaaatttt ttttttgagg ggtggg
 //

Customizing Sequence Object Construction

When you don’t need access to the complete set of annotations in a set of potentially rich sequence database entries, it is possible to configure the SeqIO parser to ignore certain sections of the sequence record. For example, let’s say you need to crunch through all of Genbank, collecting stats about the number of molecule types per species. In this case, you don’t care about the sequence or features contained in each entry. Here’s how to tell the parser to ignore these things:

my $seqin = Bio::SeqIO->new( -file => $file, -format=> 'genbank' );
my $builder = $seqin->sequence_builder();
$builder->want_all(1);
$builder->add_unwanted_slot('seq','features','annotation');

# Then go and use the SeqIO object as normal
while (my $seq = $seqin->next_seq() ) {
    # do something
}

This also skips annotations, which includes things like comments, references, and dblinks. This can speed up parsing by a factor of two or better, depending on the complexity of the sequence records.

As of version 1.5.1 this ability to customize your Sequence objects is only available for Bio::SeqIO::genbank.

See Bio::Seq::SeqBuilder and the SeqIO HOWTO for more documentation on this.

Additional Information

If you would like to learn about representing sequences and features in graphical form take a look at the Graphics HOWTO. If you have questions or comments that aren’t addressed herein then write the Bioperl community at bioperl-l@bioperl.org.

Acknowledgements

Thanks to Steven Lembark for comments and neat code discussions.