view tools/stats/aggregate_scores_in_intervals.py @ 1:cdcb0ce84a1b

Uploaded

#!/usr/bin/env python
# Greg Von Kuster
"""
usage: %prog score_file interval_file chrom start stop [out_file] [options] 
    -b, --binned: 'score_file' is actually a directory of binned array files
    -m, --mask=FILE: bed file containing regions not to consider valid
    -c, --chrom_buffer=INT: number of chromosomes (default is 3) to keep in memory when using a user supplied score file
"""

from __future__ import division
from galaxy import eggs
import pkg_resources 
pkg_resources.require( "bx-python" )
pkg_resources.require( "lrucache" )
try:
    pkg_resources.require( "python-lzo" )
except:
    pass

import psyco_full
import sys
import os, os.path
from UserDict import DictMixin
import bx.wiggle
from bx.binned_array import BinnedArray, FileBinnedArray
from bx.bitset import *
from bx.bitset_builders import *
from fpconst import isNaN
from bx.cookbook import doc_optparse
from galaxy.tools.exception_handling import *

assert sys.version_info[:2] >= ( 2, 4 )

import tempfile, struct
class PositionalScoresOnDisk:
    fmt = 'f'
    fmt_size = struct.calcsize( fmt )
    default_value = float( 'nan' )
    
    def __init__( self ):
        self.file = tempfile.TemporaryFile( 'w+b' )
        self.length = 0
    def __getitem__( self, i ):
        if i < 0: i = self.length + i
        if i < 0 or i >= self.length: return self.default_value
        try:
            self.file.seek( i * self.fmt_size )
            return struct.unpack( self.fmt, self.file.read( self.fmt_size ) )[0]
        except Exception, e:
            raise IndexError, e
    def __setitem__( self, i, value ):
        if i < 0: i = self.length + i
        if i < 0: raise IndexError, 'Negative assignment index out of range'
        if i >= self.length:
            self.file.seek( self.length * self.fmt_size )
            self.file.write( struct.pack( self.fmt, self.default_value ) * ( i - self.length ) )
            self.length = i + 1
        self.file.seek( i * self.fmt_size )
        self.file.write( struct.pack( self.fmt, value ) )
    def __len__( self ):
        return self.length
    def __repr__( self ):
        i = 0
        repr = "[ "
        for i in xrange( self.length ):
            repr = "%s %s," % ( repr, self[i] )
        return "%s ]" % ( repr )

class FileBinnedArrayDir( DictMixin ):
    """
    Adapter that makes a directory of FileBinnedArray files look like
    a regular dict of BinnedArray objects. 
    """
    def __init__( self, dir ):
        self.dir = dir
        self.cache = dict()
    def __getitem__( self, key ):
        value = None
        if key in self.cache:
            value = self.cache[key]
        else:
            fname = os.path.join( self.dir, "%s.ba" % key )
            if os.path.exists( fname ):
                value = FileBinnedArray( open( fname ) )
                self.cache[key] = value
        if value is None:
            raise KeyError( "File does not exist: " + fname )
        return value

def stop_err(msg):
    sys.stderr.write(msg)
    sys.exit()
    
def load_scores_wiggle( fname, chrom_buffer_size = 3 ):
    """
    Read a wiggle file and return a dict of BinnedArray objects keyed 
    by chromosome.
    """ 
    scores_by_chrom = dict()
    try:
        for chrom, pos, val in bx.wiggle.Reader( UCSCOutWrapper( open( fname ) ) ):
            if chrom not in scores_by_chrom:
                if chrom_buffer_size:
                    scores_by_chrom[chrom] = BinnedArray()
                    chrom_buffer_size -= 1
                else:
                    scores_by_chrom[chrom] = PositionalScoresOnDisk()
            scores_by_chrom[chrom][pos] = val
    except UCSCLimitException:
        # Wiggle data was truncated, at the very least need to warn the user.
        print 'Encountered message from UCSC: "Reached output limit of 100000 data values", so be aware your data was truncated.'
    except IndexError:
        stop_err('Data error: one or more column data values is missing in "%s"' %fname)
    except ValueError:
        stop_err('Data error: invalid data type for one or more values in "%s".' %fname)
    return scores_by_chrom

def load_scores_ba_dir( dir ):
    """
    Return a dict-like object (keyed by chromosome) that returns 
    FileBinnedArray objects created from "key.ba" files in `dir`
    """
    return FileBinnedArrayDir( dir )
    
def main():

    # Parse command line
    options, args = doc_optparse.parse( __doc__ )

    try:
        score_fname = args[0]
        interval_fname = args[1]
        chrom_col = args[2]
        start_col = args[3]
        stop_col = args[4]
        if len( args ) > 5:
            out_file = open( args[5], 'w' )
        else:
            out_file = sys.stdout
        binned = bool( options.binned )
        mask_fname = options.mask
    except:
        doc_optparse.exit()

    if score_fname == 'None':
        stop_err( 'This tool works with data from genome builds hg16, hg17 or hg18.  Click the pencil icon in your history item to set the genome build if appropriate.' )
    
    try:
        chrom_col = int(chrom_col) - 1
        start_col = int(start_col) - 1
        stop_col = int(stop_col) - 1
    except:
        stop_err( 'Chrom, start & end column not properly set, click the pencil icon in your history item to set these values.' )

    if chrom_col < 0 or start_col < 0 or stop_col < 0:
        stop_err( 'Chrom, start & end column not properly set, click the pencil icon in your history item to set these values.' )
        
    if binned:
        scores_by_chrom = load_scores_ba_dir( score_fname )
    else:
        try:
            chrom_buffer = int( options.chrom_buffer )
        except:
            chrom_buffer = 3
        scores_by_chrom = load_scores_wiggle( score_fname, chrom_buffer )

    if mask_fname:
        masks = binned_bitsets_from_file( open( mask_fname ) )
    else:
        masks = None

    skipped_lines = 0
    first_invalid_line = 0
    invalid_line = ''

    for i, line in enumerate( open( interval_fname )):
        valid = True
        line = line.rstrip('\r\n')
        if line and not line.startswith( '#' ):
            fields = line.split()
            
            try:
                chrom, start, stop = fields[chrom_col], int( fields[start_col] ), int( fields[stop_col] )
            except:
                valid = False
                skipped_lines += 1
                if not invalid_line:
                    first_invalid_line = i + 1
                    invalid_line = line
            if valid:
                total = 0
                count = 0
                min_score = 100000000
                max_score = -100000000
                for j in range( start, stop ):
                    if chrom in scores_by_chrom:
                        try:
                            # Skip if base is masked
                            if masks and chrom in masks:
                                if masks[chrom][j]:
                                    continue
                            # Get the score, only count if not 'nan'
                            score = scores_by_chrom[chrom][j]
                            if not isNaN( score ):
                                total += score
                                count += 1
                                max_score = max( score, max_score )
                                min_score = min( score, min_score )
                        except:
                            continue
                if count > 0:
                    avg = total/count
                else:
                    avg = "nan"
                    min_score = "nan"
                    max_score = "nan"
                
                # Build the resulting line of data
                out_line = []
                for k in range(0, len(fields)):
                    out_line.append(fields[k])
                out_line.append(avg)
                out_line.append(min_score)
                out_line.append(max_score)
                
                print >> out_file, "\t".join( map( str, out_line ) )
            else:
                skipped_lines += 1
                if not invalid_line:
                    first_invalid_line = i + 1
                    invalid_line = line
        elif line.startswith( '#' ):
            # We'll save the original comments
            print >> out_file, line
            
    out_file.close()

    if skipped_lines > 0:
        print 'Data issue: skipped %d invalid lines starting at line #%d which is "%s"' % ( skipped_lines, first_invalid_line, invalid_line )
        if skipped_lines == i:
            print 'Consider changing the metadata for the input dataset by clicking on the pencil icon in the history item.'

if __name__ == "__main__": main()