diff utils/gff_util.py @ 0:90100b587723 draft

Imported from capsule None
author devteam
date Tue, 01 Apr 2014 10:52:59 -0400
parents
children 8307665c4b6c
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/utils/gff_util.py	Tue Apr 01 10:52:59 2014 -0400
@@ -0,0 +1,430 @@
+"""
+Provides utilities for working with GFF files.
+"""
+
+import copy
+from bx.intervals.io import *
+from bx.tabular.io import Header, Comment
+from utils.odict import odict
+
+class GFFInterval( GenomicInterval ):
+    """
+    A GFF interval, including attributes. If file is strictly a GFF file,
+    only attribute is 'group.'
+    """
+    def __init__( self, reader, fields, chrom_col=0, feature_col=2, start_col=3, end_col=4, \
+                  strand_col=6, score_col=5, default_strand='.', fix_strand=False ):
+        # HACK: GFF format allows '.' for strand but GenomicInterval does not. To get around this,
+        # temporarily set strand and then unset after initing GenomicInterval.
+        unknown_strand = False
+        if not fix_strand and fields[ strand_col ] == '.':
+            unknown_strand = True
+            fields[ strand_col ] = '+'
+        GenomicInterval.__init__( self, reader, fields, chrom_col, start_col, end_col, strand_col, \
+                                  default_strand, fix_strand=fix_strand )
+        if unknown_strand:
+            self.strand = '.'
+            self.fields[ strand_col ] = '.'
+
+        # Handle feature, score column.
+        self.feature_col = feature_col
+        if self.feature_col >= self.nfields:
+            raise MissingFieldError( "No field for feature_col (%d)" % feature_col )
+        self.feature = self.fields[ self.feature_col ]
+        self.score_col = score_col
+        if self.score_col >= self.nfields:
+            raise MissingFieldError( "No field for score_col (%d)" % score_col )
+        self.score = self.fields[ self.score_col ]
+
+        # GFF attributes.
+        self.attributes = parse_gff_attributes( fields[8] )
+
+    def copy( self ):
+        return GFFInterval(self.reader, list( self.fields ), self.chrom_col, self.feature_col, self.start_col,
+                           self.end_col, self.strand_col, self.score_col, self.strand)
+
+class GFFFeature( GFFInterval ):
+    """
+    A GFF feature, which can include multiple intervals.
+    """
+    def __init__( self, reader, chrom_col=0, feature_col=2, start_col=3, end_col=4, \
+                  strand_col=6, score_col=5, default_strand='.', fix_strand=False, intervals=[], \
+                  raw_size=0 ):
+        # Use copy so that first interval and feature do not share fields.
+        GFFInterval.__init__( self, reader, copy.deepcopy( intervals[0].fields ), chrom_col, feature_col, \
+                              start_col, end_col, strand_col, score_col, default_strand, \
+                              fix_strand=fix_strand )
+        self.intervals = intervals
+        self.raw_size = raw_size
+        # Use intervals to set feature attributes.
+        for interval in self.intervals:
+            # Error checking. NOTE: intervals need not share the same strand.
+            if interval.chrom != self.chrom:
+                raise ValueError( "interval chrom does not match self chrom: %s != %s" % \
+                                  ( interval.chrom, self.chrom ) )
+            # Set start, end of interval.
+            if interval.start < self.start:
+                self.start = interval.start
+            if interval.end > self.end:
+                self.end = interval.end
+
+    def name( self ):
+        """ Returns feature's name. """
+        name = None
+        # Preference for name: GTF, GFF3, GFF.
+        for attr_name in [
+                           # GTF:
+                           'gene_id', 'transcript_id',
+                           # GFF3:
+                           'ID', 'id',
+                           # GFF (TODO):
+                           'group' ]:
+            name = self.attributes.get( attr_name, None )
+            if name is not None:
+                break
+        return name
+
+    def copy( self ):
+        intervals_copy = []
+        for interval in self.intervals:
+            intervals_copy.append( interval.copy() )
+        return GFFFeature(self.reader, self.chrom_col, self.feature_col, self.start_col, self.end_col, self.strand_col,
+                          self.score_col, self.strand, intervals=intervals_copy )
+
+    def lines( self ):
+        lines = []
+        for interval in self.intervals:
+            lines.append( '\t'.join( interval.fields ) )
+        return lines
+
+
+class GFFIntervalToBEDReaderWrapper( NiceReaderWrapper ):
+    """
+    Reader wrapper that reads GFF intervals/lines and automatically converts
+    them to BED format.
+    """
+
+    def parse_row( self, line ):
+        # HACK: this should return a GFF interval, but bx-python operations
+        # require GenomicInterval objects and subclasses will not work.
+        interval = GenomicInterval( self, line.split( "\t" ), self.chrom_col, self.start_col, \
+                                    self.end_col, self.strand_col, self.default_strand, \
+                                    fix_strand=self.fix_strand )
+        interval = convert_gff_coords_to_bed( interval )
+        return interval
+
+class GFFReaderWrapper( NiceReaderWrapper ):
+    """
+    Reader wrapper for GFF files.
+
+    Wrapper has two major functions:
+
+    1. group entries for GFF file (via group column), GFF3 (via id attribute),
+       or GTF (via gene_id/transcript id);
+    2. convert coordinates from GFF format--starting and ending coordinates
+       are 1-based, closed--to the 'traditional'/BED interval format--0 based,
+       half-open. This is useful when using GFF files as inputs to tools that
+       expect traditional interval format.
+    """
+
+    def __init__( self, reader, chrom_col=0, feature_col=2, start_col=3, \
+                  end_col=4, strand_col=6, score_col=5, fix_strand=False, convert_to_bed_coord=False, **kwargs ):
+        NiceReaderWrapper.__init__( self, reader, chrom_col=chrom_col, start_col=start_col, end_col=end_col, \
+                                    strand_col=strand_col, fix_strand=fix_strand, **kwargs )
+        self.feature_col = feature_col
+        self.score_col = score_col
+        self.convert_to_bed_coord = convert_to_bed_coord
+        self.last_line = None
+        self.cur_offset = 0
+        self.seed_interval = None
+        self.seed_interval_line_len = 0
+
+    def parse_row( self, line ):
+        interval = GFFInterval( self, line.split( "\t" ), self.chrom_col, self.feature_col, \
+                                self.start_col, self.end_col, self.strand_col, self.score_col, \
+                                self.default_strand, fix_strand=self.fix_strand )
+        return interval
+
+    def next( self ):
+        """ Returns next GFFFeature. """
+
+        #
+        # Helper function.
+        #
+
+        def handle_parse_error( parse_error ):
+            """ Actions to take when ParseError found. """
+            if self.outstream:
+               if self.print_delegate and hasattr(self.print_delegate,"__call__"):
+                   self.print_delegate( self.outstream, e, self )
+            self.skipped += 1
+            # no reason to stuff an entire bad file into memmory
+            if self.skipped < 10:
+               self.skipped_lines.append( ( self.linenum, self.current_line, str( e ) ) )
+
+            # For debugging, uncomment this to propogate parsing exceptions up.
+            # I.e. the underlying reason for an unexpected StopIteration exception
+            # can be found by uncommenting this.
+            # raise e
+
+        #
+        # Get next GFFFeature
+        #
+        raw_size = self.seed_interval_line_len
+
+        # If there is no seed interval, set one. Also, if there are no more
+        # intervals to read, this is where iterator dies.
+        if not self.seed_interval:
+            while not self.seed_interval:
+                try:
+                    self.seed_interval = GenomicIntervalReader.next( self )
+                except ParseError, e:
+                    handle_parse_error( e )
+                # TODO: When no longer supporting python 2.4 use finally:
+                #finally:
+                raw_size += len( self.current_line )
+
+        # If header or comment, clear seed interval and return it with its size.
+        if isinstance( self.seed_interval, ( Header, Comment ) ):
+            return_val = self.seed_interval
+            return_val.raw_size = len( self.current_line )
+            self.seed_interval = None
+            self.seed_interval_line_len = 0
+            return return_val
+
+        # Initialize feature identifier from seed.
+        feature_group = self.seed_interval.attributes.get( 'group', None ) # For GFF
+        # For GFF3
+        feature_id = self.seed_interval.attributes.get( 'ID', None )
+        feature_parent_id = self.seed_interval.attributes.get( 'Parent', None )
+        # For GTF.
+        feature_gene_id = self.seed_interval.attributes.get( 'gene_id', None )
+        feature_transcript_id = self.seed_interval.attributes.get( 'transcript_id', None )
+
+        # Read all intervals associated with seed.
+        feature_intervals = []
+        feature_intervals.append( self.seed_interval )
+        while True:
+            try:
+                interval = GenomicIntervalReader.next( self )
+                raw_size += len( self.current_line )
+            except StopIteration, e:
+                # No more intervals to read, but last feature needs to be
+                # returned.
+                interval = None
+                raw_size += len( self.current_line )
+                break
+            except ParseError, e:
+                handle_parse_error( e )
+                raw_size += len( self.current_line )
+                continue
+            # TODO: When no longer supporting python 2.4 use finally:
+            #finally:
+            #raw_size += len( self.current_line )
+
+            # Ignore comments.
+            if isinstance( interval, Comment ):
+                continue
+
+            # Determine if interval is part of feature.
+            part_of = False
+            group = interval.attributes.get( 'group', None )
+            # GFF test:
+            if group and feature_group == group:
+                part_of = True
+            # GFF3 test:
+            parent_id = interval.attributes.get( 'Parent', None )
+            cur_id = interval.attributes.get( 'ID', None )
+            if ( cur_id and cur_id == feature_id ) or ( parent_id and parent_id == feature_id ):
+                part_of = True
+            # GTF test:
+            transcript_id = interval.attributes.get( 'transcript_id', None )
+            if transcript_id and transcript_id == feature_transcript_id:
+                part_of = True
+
+            # If interval is not part of feature, clean up and break.
+            if not part_of:
+                # Adjust raw size because current line is not part of feature.
+                raw_size -= len( self.current_line )
+                break
+
+            # Interval associated with feature.
+            feature_intervals.append( interval )
+
+        # Last interval read is the seed for the next interval.
+        self.seed_interval = interval
+        self.seed_interval_line_len = len( self.current_line )
+
+        # Return feature.
+        feature = GFFFeature( self, self.chrom_col, self.feature_col, self.start_col, \
+                              self.end_col, self.strand_col, self.score_col, \
+                              self.default_strand, fix_strand=self.fix_strand, \
+                              intervals=feature_intervals, raw_size=raw_size )
+
+        # Convert to BED coords?
+        if self.convert_to_bed_coord:
+            convert_gff_coords_to_bed( feature )
+
+        return feature
+
+def convert_bed_coords_to_gff( interval ):
+    """
+    Converts an interval object's coordinates from BED format to GFF format.
+    Accepted object types include GenomicInterval and list (where the first
+    element in the list is the interval's start, and the second element is
+    the interval's end).
+    """
+    if isinstance( interval, GenomicInterval ):
+        interval.start += 1
+        if isinstance( interval, GFFFeature ):
+            for subinterval in interval.intervals:
+                convert_bed_coords_to_gff( subinterval )
+    elif type ( interval ) is list:
+        interval[ 0 ] += 1
+    return interval
+
+def convert_gff_coords_to_bed( interval ):
+    """
+    Converts an interval object's coordinates from GFF format to BED format.
+    Accepted object types include GFFFeature, GenomicInterval, and list (where
+    the first element in the list is the interval's start, and the second
+    element is the interval's end).
+    """
+    if isinstance( interval, GenomicInterval ):
+        interval.start -= 1
+        if isinstance( interval, GFFFeature ):
+            for subinterval in interval.intervals:
+                convert_gff_coords_to_bed( subinterval )
+    elif type ( interval ) is list:
+        interval[ 0 ] -= 1
+    return interval
+
+def parse_gff_attributes( attr_str ):
+    """
+    Parses a GFF/GTF attribute string and returns a dictionary of name-value
+    pairs. The general format for a GFF3 attributes string is
+
+        name1=value1;name2=value2
+
+    The general format for a GTF attribute string is
+
+        name1 "value1" ; name2 "value2"
+
+    The general format for a GFF attribute string is a single string that
+    denotes the interval's group; in this case, method returns a dictionary
+    with a single key-value pair, and key name is 'group'
+    """
+    attributes_list = attr_str.split(";")
+    attributes = {}
+    for name_value_pair in attributes_list:
+        # Try splitting by '=' (GFF3) first because spaces are allowed in GFF3
+        # attribute; next, try double quotes for GTF.
+        pair = name_value_pair.strip().split("=")
+        if len( pair ) == 1:
+            pair = name_value_pair.strip().split("\"")
+        if len( pair ) == 1:
+            # Could not split for some reason -- raise exception?
+            continue
+        if pair == '':
+            continue
+        name = pair[0].strip()
+        if name == '':
+            continue
+        # Need to strip double quote from values
+        value = pair[1].strip(" \"")
+        attributes[ name ] = value
+
+    if len( attributes ) == 0:
+        # Could not split attributes string, so entire string must be
+        # 'group' attribute. This is the case for strictly GFF files.
+        attributes['group'] = attr_str
+    return attributes
+
+def gff_attributes_to_str( attrs, gff_format ):
+    """
+    Convert GFF attributes to string. Supported formats are GFF3, GTF.
+    """
+    if gff_format == 'GTF':
+        format_string = '%s "%s"'
+        # Convert group (GFF) and ID, parent (GFF3) attributes to transcript_id, gene_id
+        id_attr = None
+        if 'group' in attrs:
+            id_attr = 'group'
+        elif 'ID' in attrs:
+            id_attr = 'ID'
+        elif 'Parent' in attrs:
+            id_attr = 'Parent'
+        if id_attr:
+            attrs['transcript_id'] = attrs['gene_id'] = attrs[id_attr]
+    elif gff_format == 'GFF3':
+        format_string = '%s=%s'
+    attrs_strs = []
+    for name, value in attrs.items():
+        attrs_strs.append( format_string % ( name, value ) )
+    return " ; ".join( attrs_strs )
+
+def read_unordered_gtf( iterator, strict=False ):
+    """
+    Returns GTF features found in an iterator. GTF lines need not be ordered
+    or clustered for reader to work. Reader returns GFFFeature objects sorted
+    by transcript_id, chrom, and start position.
+    """
+
+    # -- Get function that generates line/feature key. --
+
+    get_transcript_id = lambda fields: parse_gff_attributes( fields[8] )[ 'transcript_id' ]
+    if strict:
+        # Strict GTF parsing uses transcript_id only to group lines into feature.
+        key_fn = get_transcript_id
+    else:
+        # Use lenient parsing where chromosome + transcript_id is the key. This allows
+        # transcripts with same ID on different chromosomes; this occurs in some popular
+        # datasources, such as RefGenes in UCSC.
+        key_fn = lambda fields: fields[0] + '_' + get_transcript_id( fields )
+
+
+    # Aggregate intervals by transcript_id and collect comments.
+    feature_intervals = odict()
+    comments = []
+    for count, line in enumerate( iterator ):
+        if line.startswith( '#' ):
+            comments.append( Comment( line ) )
+            continue
+
+        line_key = key_fn( line.split('\t') )
+        if line_key in feature_intervals:
+            feature = feature_intervals[ line_key ]
+        else:
+            feature = []
+            feature_intervals[ line_key ] = feature
+        feature.append( GFFInterval( None, line.split( '\t' ) ) )
+
+    # Create features.
+    chroms_features = {}
+    for count, intervals in enumerate( feature_intervals.values() ):
+        # Sort intervals by start position.
+        intervals.sort( lambda a,b: cmp( a.start, b.start ) )
+        feature = GFFFeature( None, intervals=intervals )
+        if feature.chrom not in chroms_features:
+            chroms_features[ feature.chrom ] = []
+        chroms_features[ feature.chrom ].append( feature )
+
+    # Sort features by chrom, start position.
+    chroms_features_sorted = []
+    for chrom_features in chroms_features.values():
+        chroms_features_sorted.append( chrom_features )
+    chroms_features_sorted.sort( lambda a,b: cmp( a[0].chrom, b[0].chrom ) )
+    for features in chroms_features_sorted:
+        features.sort( lambda a,b: cmp( a.start, b.start ) )
+
+    # Yield comments first, then features.
+    # FIXME: comments can appear anywhere in file, not just the beginning.
+    # Ideally, then comments would be associated with features and output
+    # just before feature/line.
+    for comment in comments:
+        yield comment
+
+    for chrom_features in chroms_features_sorted:
+        for feature in chrom_features:
+            yield feature
+