annotate annotation_profiler_for_interval.py @ 0:3b33da018e74 draft default tip

Imported from capsule None

1 #!/usr/bin/env python
2 #Dan Blankenberg
3 #For a set of intervals, this tool returns the same set of intervals 
4 #with 2 additional fields: the name of a Table/Feature and the number of
5 #bases covered. The original intervals are repeated for each Table/Feature.
6 
7 import sys, struct, optparse, os, random
8 import bx.intervals.io
9 import bx.bitset
10 try:
11     import psyco
12     psyco.full()
13 except:
14     pass
15 
16 assert sys.version_info[:2] >= ( 2, 4 )
17 
18 class CachedRangesInFile:
19     DEFAULT_STRUCT_FORMAT = '<I'
20     def __init__( self, filename, profiler_info ):
21         self.file_size = os.stat( filename ).st_size
22         self.file = open( filename, 'rb' )
23         self.filename = filename
24         self.fmt = profiler_info.get( 'profiler_struct_format', self.DEFAULT_STRUCT_FORMAT )
25         self.fmt_size = int( profiler_info.get( 'profiler_struct_size', struct.calcsize( self.fmt ) ) )
26         self.length = int( self.file_size / self.fmt_size / 2 )
27         self._cached_ranges = [ None for i in xrange( self.length ) ]
28     def __getitem__( self, i ):
29         if self._cached_ranges[i] is not None:
30             return self._cached_ranges[i]
31         if i < 0: i = self.length + i
32         offset = i * self.fmt_size * 2
33         self.file.seek( offset )
34         try:
35             start = struct.unpack( self.fmt, self.file.read( self.fmt_size ) )[0]
36             end = struct.unpack( self.fmt, self.file.read( self.fmt_size ) )[0]
37         except Exception, e:
38             raise IndexError, e
39         self._cached_ranges[i] = ( start, end )
40         return start, end
41     def __len__( self ):
42         return self.length
43 
44 class RegionCoverage:
45     def __init__( self, filename_base, profiler_info ):
46         try:
47             self._coverage = CachedRangesInFile( "%s.covered" % filename_base, profiler_info )
48         except Exception, e:
49             #print "Error loading coverage file %s: %s" % ( "%s.covered" % filename_base, e )
50             self._coverage = []
51         try: 
52             self._total_coverage = int( open( "%s.total_coverage" % filename_base ).read() )
53         except Exception, e:
54             #print "Error loading total coverage file %s: %s" % ( "%s.total_coverage" % filename_base, e )
55             self._total_coverage = 0
56     def get_start_index( self, start ):
57         #binary search: returns index of range closest to start
58         if start > self._coverage[-1][1]:
59             return len( self._coverage ) - 1
60         i = 0
61         j = len( self._coverage) - 1
62         while i < j:
63             k = ( i + j ) / 2
64             if start <= self._coverage[k][1]:
65                 j = k
66             else:
67                 i = k + 1
68         return i
69     def get_coverage( self, start, end ):
70         return self.get_coverage_regions_overlap( start, end )[0]
71     def get_coverage_regions_overlap( self, start, end ):
72         return self.get_coverage_regions_index_overlap( start, end )[0:2]
73     def get_coverage_regions_index_overlap( self, start, end ):
74         if len( self._coverage ) < 1 or start > self._coverage[-1][1] or end < self._coverage[0][0]:
75             return 0, 0, 0
76         if self._total_coverage and start <= self._coverage[0][0] and end >= self._coverage[-1][1]:
77             return self._total_coverage, len( self._coverage ), 0
78         coverage = 0
79         region_count = 0
80         start_index = self.get_start_index( start )
81         for i in xrange( start_index, len( self._coverage ) ):
82             c_start, c_end = self._coverage[i]
83             if c_start > end:
84                 break
85             if c_start <= end and c_end >= start:
86                 coverage += min( end, c_end ) - max( start, c_start )
87                 region_count += 1
88         return coverage, region_count, start_index
89 
90 class CachedCoverageReader:
91     def __init__( self, base_file_path, buffer = 10, table_names = None, profiler_info = None ):
92         self._base_file_path = base_file_path
93         self._buffer = buffer #number of chromosomes to keep in memory at a time
94         self._coverage = {}
95         if table_names is None: table_names = [ table_dir for table_dir in os.listdir( self._base_file_path ) if os.path.isdir( os.path.join( self._base_file_path, table_dir ) ) ]
96         for tablename in table_names: self._coverage[tablename] = {}
97         if profiler_info is None: profiler_info = {}
98         self._profiler_info = profiler_info
99     def iter_table_coverage_by_region( self, chrom, start, end ):
00         for tablename, coverage, regions in self.iter_table_coverage_regions_by_region( chrom, start, end ):
01             yield tablename, coverage
02     def iter_table_coverage_regions_by_region( self, chrom, start, end ):
03         for tablename, coverage, regions, index in self.iter_table_coverage_regions_index_by_region( chrom, start, end ):
04             yield tablename, coverage, regions
05     def iter_table_coverage_regions_index_by_region( self, chrom, start, end ):
06         for tablename, chromosomes in self._coverage.iteritems():
07             if chrom not in chromosomes:
08                 if len( chromosomes ) >= self._buffer:
09                     #randomly remove one chromosome from this table
10                     del chromosomes[ chromosomes.keys().pop( random.randint( 0, self._buffer - 1 ) ) ]
11                 chromosomes[chrom] = RegionCoverage( os.path.join ( self._base_file_path, tablename, chrom ), self._profiler_info )
12             coverage, regions, index = chromosomes[chrom].get_coverage_regions_index_overlap( start, end )
13             yield tablename, coverage, regions, index
14 
15 class TableCoverageSummary:
16     def __init__( self, coverage_reader, chrom_lengths ):
17         self.coverage_reader = coverage_reader
18         self.chrom_lengths = chrom_lengths
19         self.chromosome_coverage = {} #dict of bitset by chromosome holding user's collapsed input intervals
20         self.total_interval_size = 0 #total size of user's input intervals
21         self.total_interval_count = 0 #total number of user's input intervals
22         self.table_coverage = {} #dict of total coverage by user's input intervals by table
23         self.table_chromosome_size = {} #dict of dict of table:chrom containing total coverage of table for a chrom
24         self.table_chromosome_count = {} #dict of dict of table:chrom containing total number of coverage ranges of table for a chrom
25         self.table_regions_overlaped_count = {} #total number of table regions overlaping user's input intervals (non unique)
26         self.interval_table_overlap_count = {} #total number of user input intervals which overlap table
27         self.region_size_errors = {} #dictionary of lists of invalid ranges by chromosome
28     def add_region( self, chrom, start, end ):
29         chrom_length = self.chrom_lengths.get( chrom )
30         region_start = min( start, chrom_length )
31         region_end = min( end, chrom_length )
32         region_length = region_end - region_start
33         
34         if region_length < 1 or region_start != start or region_end != end:
35             if chrom not in self.region_size_errors:
36                 self.region_size_errors[chrom] = []
37             self.region_size_errors[chrom].append( ( start, end ) )
38             if region_length < 1: return
39         
40         self.total_interval_size += region_length
41         self.total_interval_count += 1
42         if chrom not in self.chromosome_coverage:
43             self.chromosome_coverage[chrom] = bx.bitset.BitSet( chrom_length )
44         
45         self.chromosome_coverage[chrom].set_range( region_start, region_length )
46         for table_name, coverage, regions in self.coverage_reader.iter_table_coverage_regions_by_region( chrom, region_start, region_end ):
47             if table_name not in self.table_coverage:
48                 self.table_coverage[table_name] = 0
49                 self.table_chromosome_size[table_name] = {}
50                 self.table_regions_overlaped_count[table_name] = 0
51                 self.interval_table_overlap_count[table_name] = 0
52                 self.table_chromosome_count[table_name] = {}
53             if chrom not in self.table_chromosome_size[table_name]:
54                 self.table_chromosome_size[table_name][chrom] = self.coverage_reader._coverage[table_name][chrom]._total_coverage
55                 self.table_chromosome_count[table_name][chrom] = len( self.coverage_reader._coverage[table_name][chrom]._coverage )
56             self.table_coverage[table_name] += coverage
57             if coverage:
58                 self.interval_table_overlap_count[table_name] += 1
59             self.table_regions_overlaped_count[table_name] += regions
60     def iter_table_coverage( self ):
61         def get_nr_coverage():
62             #returns non-redundant coverage, where user's input intervals have been collapse to resolve overlaps
63             table_coverage = {} #dictionary of tables containing number of table bases overlaped by nr intervals
64             interval_table_overlap_count = {} #dictionary of tables containing number of nr intervals overlaping table
65             table_regions_overlap_count = {} #dictionary of tables containing number of regions overlaped (unique)
66             interval_count = 0 #total number of nr intervals
67             interval_size = 0 #holds total size of nr intervals
68             region_start_end = {} #holds absolute start,end for each user input chromosome
69             for chrom, chromosome_bitset in self.chromosome_coverage.iteritems():
70                 #loop through user's collapsed input intervals
71                 end = 0
72                 last_end_index = {}
73                 interval_size += chromosome_bitset.count_range()
74                 while True:
75                     if end >= chromosome_bitset.size: break
76                     start = chromosome_bitset.next_set( end )
77                     if start >= chromosome_bitset.size: break
78                     end = chromosome_bitset.next_clear( start )
79                     interval_count += 1
80                     if chrom not in region_start_end:
81                         region_start_end[chrom] = [start, end]
82                     else:
83                         region_start_end[chrom][1] = end
84                     for table_name, coverage, region_count, start_index in self.coverage_reader.iter_table_coverage_regions_index_by_region( chrom, start, end ):
85                         if table_name not in table_coverage:
86                             table_coverage[table_name] = 0
87                             interval_table_overlap_count[table_name] = 0
88                             table_regions_overlap_count[table_name] = 0
89                         table_coverage[table_name] += coverage
90                         if coverage:
91                             interval_table_overlap_count[table_name] += 1
92                             table_regions_overlap_count[table_name] += region_count
93                             if table_name in last_end_index and last_end_index[table_name] == start_index:
94                                 table_regions_overlap_count[table_name] -= 1
95                             last_end_index[table_name] = start_index + region_count - 1
96             table_region_coverage = {} #total coverage for tables by bounding nr interval region
97             table_region_count = {} #total number for tables by bounding nr interval region
98             for chrom, start_end in region_start_end.items():
99                 for table_name, coverage, region_count in self.coverage_reader.iter_table_coverage_regions_by_region( chrom, start_end[0], start_end[1] ):
00                     if table_name not in table_region_coverage:
01                         table_region_coverage[table_name] = 0
02                         table_region_count[table_name] = 0
03                     table_region_coverage[table_name] += coverage
04                     table_region_count[table_name] += region_count
05             return table_region_coverage, table_region_count, interval_count, interval_size, table_coverage, table_regions_overlap_count, interval_table_overlap_count
06         table_region_coverage, table_region_count, nr_interval_count, nr_interval_size, nr_table_coverage, nr_table_regions_overlap_count, nr_interval_table_overlap_count = get_nr_coverage()
07         for table_name in self.table_coverage:
08             #TODO: determine a type of statistic, then calculate and report here
09             yield table_name, sum( self.table_chromosome_size.get( table_name, {} ).values() ), sum( self.table_chromosome_count.get( table_name, {} ).values() ), table_region_coverage.get( table_name, 0 ), table_region_count.get( table_name, 0 ), self.total_interval_count, self.total_interval_size,  self.table_coverage[table_name], self.table_regions_overlaped_count.get( table_name, 0), self.interval_table_overlap_count.get( table_name, 0 ), nr_interval_count, nr_interval_size, nr_table_coverage[table_name], nr_table_regions_overlap_count.get( table_name, 0 ), nr_interval_table_overlap_count.get( table_name, 0 )
10 
11 def profile_per_interval( interval_filename, chrom_col, start_col, end_col, out_filename, keep_empty, coverage_reader ):
12     out = open( out_filename, 'wb' )
13     for region in bx.intervals.io.NiceReaderWrapper( open( interval_filename, 'rb' ), chrom_col = chrom_col, start_col = start_col, end_col = end_col, fix_strand = True, return_header = False, return_comments = False ):
14         for table_name, coverage, region_count in coverage_reader.iter_table_coverage_regions_by_region( region.chrom, region.start, region.end ):
15             if keep_empty or coverage:
16                 #only output regions that have atleast 1 base covered unless empty are requested
17                 out.write( "%s\t%s\t%s\t%s\n" % ( "\t".join( region.fields ), table_name, coverage, region_count ) )
18     out.close()
19 
20 def profile_summary( interval_filename, chrom_col, start_col, end_col, out_filename, keep_empty, coverage_reader, chrom_lengths ):
21     out = open( out_filename, 'wb' )
22     table_coverage_summary = TableCoverageSummary( coverage_reader, chrom_lengths )
23     for region in bx.intervals.io.NiceReaderWrapper( open( interval_filename, 'rb' ), chrom_col = chrom_col, start_col = start_col, end_col = end_col, fix_strand = True, return_header = False, return_comments = False ):
24         table_coverage_summary.add_region( region.chrom, region.start, region.end )
25     
26     out.write( "#tableName\ttableChromosomeCoverage\ttableChromosomeCount\ttableRegionCoverage\ttableRegionCount\tallIntervalCount\tallIntervalSize\tallCoverage\tallTableRegionsOverlaped\tallIntervalsOverlapingTable\tnrIntervalCount\tnrIntervalSize\tnrCoverage\tnrTableRegionsOverlaped\tnrIntervalsOverlapingTable\n" )
27     for table_name, table_chromosome_size, table_chromosome_count, table_region_coverage, table_region_count, total_interval_count, total_interval_size, total_coverage, table_regions_overlaped_count, interval_region_overlap_count, nr_interval_count, nr_interval_size, nr_coverage, nr_table_regions_overlaped_count, nr_interval_table_overlap_count in table_coverage_summary.iter_table_coverage():
28         if keep_empty or total_coverage:
29             #only output tables that have atleast 1 base covered unless empty are requested
30             out.write( "%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\t%s\n" % ( table_name, table_chromosome_size, table_chromosome_count, table_region_coverage, table_region_count, total_interval_count, total_interval_size, total_coverage, table_regions_overlaped_count, interval_region_overlap_count, nr_interval_count, nr_interval_size, nr_coverage, nr_table_regions_overlaped_count, nr_interval_table_overlap_count ) )
31     out.close()
32     
33     #report chrom size errors as needed:
34     if table_coverage_summary.region_size_errors:
35         print "Regions provided extended beyond known chromosome lengths, and have been truncated as necessary, for the following intervals:"
36         for chrom, regions in table_coverage_summary.region_size_errors.items():
37             if len( regions ) > 3:
38                 extra_region_info = ", ... "
39             else:
40                 extra_region_info = ""
41             print "%s has max length of %s, exceeded by %s%s." % ( chrom, chrom_lengths.get( chrom ), ", ".join( map( str, regions[:3] ) ), extra_region_info )
42 
43 class ChromosomeLengths:
44     def __init__( self, profiler_info ):
45         self.chroms = {}
46         self.default_bitset_size = int( profiler_info.get( 'bitset_size', bx.bitset.MAX ) )
47         chroms = profiler_info.get( 'chromosomes', None )
48         if chroms:
49             for chrom in chroms.split( ',' ):
50                 for fields in chrom.rsplit( '=', 1 ):
51                     if len( fields ) == 2:
52                         self.chroms[ fields[0] ] = int( fields[1] )
53                     else:
54                         self.chroms[ fields[0] ] = self.default_bitset_size
55     def get( self, name ):
56         return self.chroms.get( name, self.default_bitset_size )
57 
58 def parse_profiler_info( filename ):
59     profiler_info = {}
60     try:
61         for line in open( filename ):
62             fields = line.rstrip( '\n\r' ).split( '\t', 1 )
63             if len( fields ) == 2:
64                 if fields[0] in profiler_info:
65                     if not isinstance( profiler_info[ fields[0] ], list ):
66                         profiler_info[ fields[0] ] = [ profiler_info[ fields[0] ] ]
67                     profiler_info[ fields[0] ].append( fields[1] )
68                 else:
69                     profiler_info[ fields[0] ] = fields[1]
70     except:
71         pass #likely missing file
72     return profiler_info
73 
74 def __main__():
75     parser = optparse.OptionParser()
76     parser.add_option(
77         '-k','--keep_empty',
78         action="store_true",
79         dest='keep_empty',
80         default=False,
81         help='Keep tables with 0 coverage'
82     )
83     parser.add_option(
84         '-b','--buffer',
85         dest='buffer',
86         type='int',default=10,
87         help='Number of Chromosomes to keep buffered'
88     )
89     parser.add_option(
90         '-c','--chrom_col',
91         dest='chrom_col',
92         type='int',default=1,
93         help='Chromosome column'
94     )
95     parser.add_option(
96         '-s','--start_col',
97         dest='start_col',
98         type='int',default=2,
99         help='Start Column'
00     )
01     parser.add_option(
02         '-e','--end_col',
03         dest='end_col',
04         type='int',default=3,
05         help='End Column'
06     )
07     parser.add_option(
08         '-p','--path',
09         dest='path',
10         type='str',default='/galaxy/data/annotation_profiler/hg18',
11         help='Path to profiled data for this organism'
12     )
13     parser.add_option(
14         '-t','--table_names',
15         dest='table_names',
16         type='str',default='None',
17         help='Table names requested'
18     )
19     parser.add_option(
20         '-i','--input',
21         dest='interval_filename',
22         type='str',
23         help='Input Interval File'
24     )
25     parser.add_option(
26         '-o','--output',
27         dest='out_filename',
28         type='str',
29         help='Input Interval File'
30     )
31     parser.add_option(
32         '-S','--summary',
33         action="store_true",
34         dest='summary',
35         default=False,
36         help='Display Summary Results'
37     )
38     
39     options, args = parser.parse_args()
40     
41     assert os.path.isdir( options.path ), IOError( "Configuration error: Table directory is missing (%s)" % options.path )
42     
43     #get profiler_info
44     profiler_info = parse_profiler_info( os.path.join( options.path, 'profiler_info.txt' ) )
45     
46     table_names = options.table_names.split( "," )
47     if table_names == ['None']: table_names = None
48     coverage_reader = CachedCoverageReader( options.path, buffer = options.buffer, table_names = table_names, profiler_info = profiler_info )
49     
50     if options.summary:
51         profile_summary( options.interval_filename, options.chrom_col - 1, options.start_col - 1, options.end_col -1, options.out_filename, options.keep_empty, coverage_reader, ChromosomeLengths( profiler_info ) )
52     else:
53         profile_per_interval( options.interval_filename, options.chrom_col - 1, options.start_col - 1, options.end_col -1, options.out_filename, options.keep_empty, coverage_reader )
54     
55     #print out data version info
56     print 'Data version (%s:%s:%s)' % ( profiler_info.get( 'dbkey', 'unknown' ), profiler_info.get( 'profiler_hash', 'unknown' ), profiler_info.get( 'dump_time', 'unknown' ) )
57 
58 if __name__ == "__main__": __main__()