def main(): """" Set up the CLI """ parser = argparse.ArgumentParser(description=__doc__) parser.add_argument("input_files", nargs="+", default=[], metavar="INFILE", help="List of hit tables to process") parser.add_argument("-o", "--outfile", dest="outfile", metavar="OUTFILE", help="Write count table to OUTFILE") parser.add_argument("-r", "--rank", dest="ranks", default=None, metavar="RANK", action="append", help=""" Rank(s) to collect counts on. Use flag multiple times to specify multiple ranks. If multiple values given, one table produced for each with rank name appended to file name. Defaults to all major ranks between phylum and species. Corresponds to rank names in nodes.dmp. To see list run: 'cut -f5 nodes.dmp | uniq | sort | uniq' in ncbi tax dir. Will also accept 'organism' to mean no rank (ie, just the organism name).""") parser.add_argument( "-s", "--collapseToDomain", default=False, action="store_true", help="Collapse all taxa below given rank down to " "superkingdom/domain. EG: in the genus output, anything " "assigned to Cyanobactia, will be lumped in with all " "other bacteria") parser.add_argument( "--proportional", dest="proportional", default=False, action="store_true", help="""When using tophit or toporg, redistribute proportionally instead of winner take all""") parser.add_argument( "-R", "--printRank", dest="printRanks", action="append", help="Include indeicated rank(s) in lineage of printed taxa. " "Will be ignored if beyond the rank of the taxa " "(IE We can't include species if the taxon being counted " "is genus)") # option for deconvoluting clusters or assemblies add_weight_arguments(parser, multiple=True) # cutoff options add_count_arguments(parser) # format, tax dir, and more add_taxon_arguments( parser, choices={ 'countMethod': ( 'LCA', 'all', 'first', 'most', 'tophit', 'toporg', 'consensus')}) # log level and help add_universal_arguments(parser) arguments = parser.parse_args() setup_logging(arguments) if arguments.proportional and \ arguments.countMethod not in ['tophit', 'toporg']: parser.error("--proportinal only has meaning " "if using tophit or toporg") if len(arguments.input_files) == 0: parser.error("Must supply at least one m8 file to parse") # Handle the case where Galaxy tries to set None as a string arguments.ranks = checkNoneOption(arguments.ranks) arguments.printRanks = checkNoneOption(arguments.printRanks) logging.info("Printing out ranks: %r", arguments.ranks) # Set defaults and check for some conflicts if arguments.ranks is None and arguments.taxdir is None: # using hit names only arguments.ranks = [ORG_RANK] if arguments.printRanks is not None: parser.error("Display ranks are not used without taxonomic info") else: if arguments.taxdir is None: parser.error("Cannot select ranks without a taxonomy") if arguments.ranks is None: # set a default arguments.ranks = [ 'phylum', 'class', 'order', 'family', 'genus', 'species'] try: # Make sure the rank lists make sense arguments.ranks = cleanRanks(arguments.ranks) if arguments.printRanks is not None: arguments.printRanks = cleanRanks(arguments.printRanks) except Exception as e: parser.error(str(e)) # load weights file sequenceWeights = loadSequenceWeights(arguments.weights) # only print to stdout if there is a single rank if len(arguments.ranks) > 1 and arguments.outfile is None: parser.error("STDOUT only works if a single rank is chosen!") # Because rank is used in parsing hits, we can only do multiple ranks for # certain kinds of count methods if len(arguments.ranks) > 1: rank = None if arguments.countMethod in ['consensus', 'most']: parser.error( "Using multiple ranks does not work with the 'consensus' " "or 'most' counting methods. LCA should give the same " "results as consensus. If you really want to do this, " "use a bash loop:'for rank in phylum order genus; do " "COMMAND -r ${rank}; done'") else: rank = arguments.ranks[0] # load necessary maps (taxonomy, hitStringMap) = readMaps(arguments) # parse input files fileCounts = {} totals = {} fileLabels = {} sortedLabels = [] # Allow for file names to be preceded with TAG= for filename in arguments.input_files: bits = filename.split("=", 1) if len(bits) > 1: (filetag, filename) = bits else: filetag = filename fileLabels[filename] = filetag # keep order so that column order matches arguments sortedLabels.append(filetag) fileCounts[filetag] = {} totals[filetag] = 0 params = FilterParams.create_from_arguments(arguments) if arguments.countMethod == 'tophit' or arguments.countMethod == 'toporg': # Process all files at once and use overall abundance to pick best hits from edl import redistribute multifile = redistribute.multipleFileWrapper(fileLabels.keys()) if arguments.countMethod == 'tophit': # don't give any taxonomy, just map to accessions for # redistribution readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, winnerTakeAll=not arguments.proportional, parseStyle=arguments.parseStyle, sequenceWeights=sequenceWeights) # define method to turn Hits into orgnaisms hitTranslator = getHitTranslator(parseStyle=arguments.parseStyle, taxonomy=taxonomy, hitStringMap=hitStringMap) translateHit = lambda hit: hitTranslator.translateHit(hit=hit)[0] else: # translate to organism before finding most abundant readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, returnTranslations=True, winnerTakeAll=not arguments.proportional, taxonomy=taxonomy, hitStringMap=hitStringMap, parseStyle=ACCS) # Organisms will be returned, make translator trivial: translateHit = passThrough # use read->file mapping and hit translator to get file based counts # from returned (read,Hit) pairs increment = 1 for (read_name, hit) in readHits: file_name, read_name = read_name.split("/", 1) file_tag = fileLabels[unquote_plus(file_name)] taxon = translateHit(hit) taxcount = fileCounts[file_tag].setdefault(taxon, 0) if sequenceWeights is not None: increment = sequenceWeights.get(read_name, 1) fileCounts[file_tag][taxon] = taxcount + increment totals[file_tag] += increment logging.debug(str(totals)) else: # Original way, just process each file separately for (filename, filetag) in fileLabels.items(): infile = open(filename, 'rU') hitIter = parseM8FileIter(infile, hitStringMap, params, arguments.parseStyle, arguments.countMethod, taxonomy=taxonomy, rank=rank) (total, counts, hitMap) = \ countIterHits(hitIter, allMethod=arguments.allMethod, weights=sequenceWeights) fileCounts[filetag] = counts totals[filetag] = total logging.info( "parsed %d hits (%d unique) for %d reads from %s", total, len(counts), len(hitMap), filename) infile.close() printCountTablesByRank(fileCounts, totals, sortedLabels, arguments)
def pickBestHitByAbundance(m8stream, filterParams=None, return_lines=True, return_translations=False, organismCounts=None, winnerTakeAll=False, sequenceWeights=None, **kwargs): """ Given a hit table with (potentially) multiple hits for each read. Select the best hit for each read. Hits are parsed from given hit table (m8stream) if given a FilterParams object, otherwise it is assumed that m8stream is an iterator over Hit objects. Remaining keyword arguments are used to translate hits to accessions, organisms, or anything else using a HitTranslator. Ambiguous hits (multiple 'best' hits to one read) are resolved as follows: given a set of reads that all hit the same list of translated hits: divvy up reads so that the abundance ratios change minimally Abundance is recorded for whatever the Hittranslator returns. If a hit map and taxonomy are given, this will be organisms, if only the parseStyle is given and it's set to ACC, then accessions will be the currency. The default is HITID. Yields (read,hit) tuples, (read, [translated hits]) tuples, or hit table lines. """ if return_lines and return_translations: return_lines = False logger.warn("return_translations overrides return_lines!") # filtered hits if filterParams is None: hitIter = m8stream else: hitIter = filterM8Stream(m8stream, filterParams, return_lines=False) # custom function for pulling orgs from hits # if no settings given, just use the hit ID as the 'organism' kwargs.setdefault("parseStyle", HITID) hitTranslator = getHitTranslator(**kwargs) # we need to keep track of lots of things orgCounts = {} totalReads = 0 unambiguousReads = 0 ambiguousReads = 0 sameOrgCount = 0 ambiguousHits = {} # Check to see if organism counts were given if organismCounts is not None: if isinstance(organismCounts, str): organismCounts = getOrganismCountsFromFile(organismCounts) # loop over hits and yield unambiguous ones # Save ambiguous hits and org abundances logger.debug(str(hitIter)) for (read, hits) in hitIter: logger.debug("Read: %s" % (read)) totalReads += 1 hitByOrg = {} orgs = [] count = 0 for hit in hits: count += 1 hitOrgs = hitTranslator.translateHit(hit) logger.debug("Hit: %s (%s), %s" % (hit.hit, hitOrgs, hit.score)) orgs.extend(hitOrgs) for org in hitOrgs: if org in hitByOrg: # This should be REALLY rare. sameOrgCount += 1 sameOrgExample = (read, hit.hit, org) logger.warn( "Read (%s) has two best hits to same org (%s)!" % (read, org)) # always keep the first alphabetically, for reproducibility if hit.hit < hitByOrg[org].hit: hitByOrg[org] = hit else: hitByOrg[org] = hit orgs = tuple(sorted(set(orgs))) if count == 0: # This *should* never happen logger.error("No hits for %s!!!!!" % (read)) raise Exception("Read (%s) has not hits. This shouldn't happen." % (read)) elif count == 1 or len(hitByOrg) == 1: logger.debug("Read is UNambiguous") unambiguousReads += 1 for org in orgs: if sequenceWeights is not None: increment = sequenceWeights.get(read, 1) else: increment = 1 orgCounts[org] = orgCounts.get(org, 0) + increment if return_lines: yield hit.line elif return_translations: yield (read, orgs) else: yield (read, hit) else: logger.debug("Read IS ambiguous") ambiguousReads += 1 if organismCounts is None: # If we don't have count data to start, save these til the end ambiguousHits.setdefault(orgs, []).append(hitByOrg) else: # Use given counts to resolve for (hit, org) in assignHits(orgs, [ hitByOrg, ], organismCounts, winnerTakeAll): yield formatReturn(hit, org, return_lines, return_translations) logger.info("Processed %d reads:" % (totalReads)) logger.info("Collected unambiguous counts for %d orgs from %d reads" % (len(orgCounts), unambiguousReads)) # if we used given organism counts, then we are done if organismCounts is not None: return # otherwise, we have ambiguous reads to resolve logger.info("Need to resolve %d ambiguous reads hitting %d orgs" % (ambiguousReads, len(ambiguousHits))) if sameOrgCount > 0: elements = list(sameOrgExample) elements.insert(0, sameOrgCount) logger.warn("found %d cases where a read had an extra hit to the same " "organism. For Example: %s (%s,%s)" % tuple(elements)) # loop over ambiguous hits (grouped by possible orgs) and pick one for # each read ambiguousReads = 0 # for orgs, hits in ambiguousHits.items(): for orgs in sorted(ambiguousHits.keys()): hits = ambiguousHits[orgs] for (hit, org) in assignHits(orgs, hits, orgCounts, winnerTakeAll): ambiguousReads += 1 yield formatReturn(hit, org, return_lines, return_translations) logger.info("Selected top hit for %d ambiguous reads for a total of %d " "returned hit assignments" % (ambiguousReads, ambiguousReads + unambiguousReads))
def pickBestHitByAbundance(m8stream, filterParams=None, returnLines=True, returnTranslations=False, organismCounts=None, winnerTakeAll=False, sequenceWeights=None, **kwargs): """ Given a hit table with (potentially) multiple hits for each read. Select the best hit for each read. Hits are parsed from given hit table (m8stream) if given a FilterParams object, otherwise it is assumed that m8stream is an iterator over Hit objects. Remaining keyword arguments are used to translate hits to accessions, organisms, or anything else using a HitTranslator. Ambiguous hits (multiple 'best' hits to one read) are resolved as follows: given a set of reads that all hit the same list of translated hits: divvy up reads so that the abundance ratios change minimally Abundance is recorded for whatever the Hittranslator returns. If a hit map and taxonomy are given, this will be organisms, if only the parseStyle is given and it's set to ACC, then accessions will be the currency. The default is HITID. Yields (read,hit) tuples, (read, [translated hits]) tuples, or hit table lines. """ if returnLines and returnTranslations: returnLines = False logger.warn("returnTranslations overrides returnLines!") # filtered hits if filterParams is None: hitIter = m8stream else: hitIter = blastm8.filterM8Stream( m8stream, filterParams, returnLines=False) # custom function for pulling orgs from hits # if no settings given, just use the hit ID as the 'organism' kwargs.setdefault("parseStyle", HITID) hitTranslator = getHitTranslator(**kwargs) # we need to keep track of lots of things orgCounts = {} totalReads = 0 unambiguousReads = 0 ambiguousReads = 0 sameOrgCount = 0 ambiguousHits = {} # Check to see if organism counts were given if organismCounts is not None: if isinstance(organismCounts, str): organismCounts = getOrganismCountsFromFile(organismCounts) # loop over hits and yield unambiguous ones # Save ambiguous hits and org abundances logger.debug(str(hitIter)) for (read, hits) in hitIter: logger.debug("Read: %s" % (read)) totalReads += 1 hitByOrg = {} orgs = [] count = 0 for hit in hits: count += 1 hitOrgs = hitTranslator.translateHit(hit) logger.debug("Hit: %s (%s), %s" % (hit.hit, hitOrgs, hit.score)) orgs.extend(hitOrgs) for org in hitOrgs: if org in hitByOrg: # This should be REALLY rare. sameOrgCount += 1 sameOrgExample = (read, hit.hit, org) logger.warn( "Read (%s) has two best hits to same org (%s)!" % (read, org)) # always keep the first alphabetically, for reproducibility if hit.hit < hitByOrg[org].hit: hitByOrg[org] = hit else: hitByOrg[org] = hit orgs = tuple(sorted(set(orgs))) if count == 0: # This *should* never happen logger.error("No hits for %s!!!!!" % (read)) raise Exception( "Read (%s) has not hits. This shouldn't happen." % (read)) elif count == 1 or len(hitByOrg) == 1: logger.debug("Read is UNambiguous") unambiguousReads += 1 for org in orgs: if sequenceWeights is not None: increment = sequenceWeights.get(read, 1) else: increment = 1 orgCounts[org] = orgCounts.get(org, 0) + increment if returnLines: yield hit.line elif returnTranslations: yield (read, orgs) else: yield (read, hit) else: logger.debug("Read IS ambiguous") ambiguousReads += 1 if organismCounts is None: # If we don't have count data to start, save these til the end ambiguousHits.setdefault(orgs, []).append(hitByOrg) else: # Use given counts to resolve for ( hit, org) in assignHits( orgs, [ hitByOrg, ], organismCounts, winnerTakeAll): yield formatReturn(hit, org, returnLines, returnTranslations) logger.info("Processed %d reads:" % (totalReads)) logger.info( "Collected unambiguous counts for %d orgs from %d reads" % (len(orgCounts), unambiguousReads)) # if we used given organism counts, then we are done if organismCounts is not None: return # otherwise, we have ambiguous reads to resolve logger.info( "Need to resolve %d ambiguous reads hitting %d orgs" % (ambiguousReads, len(ambiguousHits))) if sameOrgCount > 0: elements = list(sameOrgExample) elements.insert(0, sameOrgCount) logger.warn( "found %d cases where a read had an extra hit to the same " "organism. For Example: %s (%s,%s)" % tuple(elements)) # loop over ambiguous hits (grouped by possible orgs) and pick one for # each read ambiguousReads = 0 # for orgs, hits in ambiguousHits.items(): for orgs in sorted(ambiguousHits.keys()): hits = ambiguousHits[orgs] for (hit, org) in assignHits(orgs, hits, orgCounts, winnerTakeAll): ambiguousReads += 1 yield formatReturn(hit, org, returnLines, returnTranslations) logger.info( "Selected top hit for %d ambiguous reads for a total of %d " "returned hit assignments" % (ambiguousReads, ambiguousReads + unambiguousReads))
def main(): description = __doc__ parser = argparse.ArgumentParser(description=description) parser.add_argument("input_files", nargs="+", default=[], metavar="INFILE", help="List of hit tables to process") parser.add_argument("-o", "--outfile", dest="output_file", metavar="OUTFILE", help="Write count table to OUTFILE") parser.add_argument("-l", "--level", dest="levels", default=None, metavar="LEVEL", action="append", help=""" Level(s) to collect counts on. Use flag multiple times to specify multiple levels. If multiple values given, one table produced for each with rank name appended to file name. Levels can be an integer (1-3) for KEGG or SEED levels, any one of 'gene', 'role', 'family', 'ko', or 'ortholog' (which are all synonyms), or anything not synonymous with 'gene' to get CAZy groups. Defaults to ortholog/role and levels 1, 2, and 3 for KEGG and SEED and gene and group for CAZy and COG.""") # option for deconvoluting clusters or assemblies add_weight_arguments(parser, multiple=True) # cutoff options add_count_arguments(parser) # format, ortholog heirarchy, and more kegg.add_path_arguments( parser, defaults={'countMethod': 'tophit'}, choices={'countMethod': ('tophit', 'first', 'most', 'all', 'consensus')}, helps={'countMethod': ("How to deal with counts from multiple hits. ('first': " "just use the first hit, 'most': " "can return multiple hits, 'all': return every hit, " "consensus: return None unless all the same). Do not " "use most or consensus with more than one level at a time. " "Default is 'tophit': This breaks any ties by choosing " "the most abundant hit based on other unambiguous " "assignments.")}) # log level and help add_universal_arguments(parser) arguments = parser.parse_args() setup_logging(arguments) if len(arguments.input_files) == 0: parser.error("Must supply at least one m8 file to parse") # Set defaults and check for some conflicts if arguments.levels is None and arguments.heirarchyFile is None: # using hit names only arguments.levels = [None] else: if arguments.heirarchyFile is None \ and arguments.heirarchyType != 'cazy': logging.warning("Type: %s", arguments.heirarchyType) parser.error("Cannot select levels without a heirarchy (ko) file") if arguments.levels is None: # set a default if arguments.heirarchyType is 'kegg': arguments.levels = ['ko', '1', '2', 'pathway'] if arguments.heirarchyType is 'seed': arguments.levels = ['role', '1', '2', 'subsystem'] else: arguments.levels = ['gene', 'group'] try: # Make sure the rank lists make sense arguments.levels = cleanLevels(arguments.levels) except Exception as e: parser.error(str(e)) # load weights file sequenceWeights = loadSequenceWeights(arguments.weights) # only print to stdout if there is a single level if len(arguments.levels) > 1 and arguments.output_file is None: parser.error("STDOUT only works if a single level is chosen!") cutoff = arguments.cutoff # map reads to hits if arguments.mapFile is not None: if arguments.mapStyle == 'auto': with open(arguments.mapFile) as f: firstLine = next(f) while len(firstLine) == 0 or firstLine[0] == '#': firstLine = next(f) if koMapRE.search(firstLine): arguments.mapStyle = 'kegg' elif seedMapRE.search(firstLine): arguments.mapStyle = 'seed' elif tabMapRE.search(firstLine): arguments.mapStyle = 'tab' # elif cogMapRE.search(firstLine): # arguments.mapStyle='cog' else: raise Exception( "Cannot figure out map type from first line:\n%s" % (firstLine)) logging.info("Map file seems to be: %s", arguments.mapStyle) if arguments.mapStyle == 'kegg': valueMap = kegg.parseLinkFile(arguments.mapFile) elif arguments.mapStyle == 'seed': valueMap = kegg.parseSeedMap(arguments.mapFile) # elif arguments.mapStyle=='cog': # valueMap=kegg.parseCogMap(arguments.mapFile) else: if arguments.parseStyle == GIS: keyType = int else: keyType = None valueMap = parseMapFile( arguments.mapFile, valueType=None, valueDelim=arguments.tab_map_delim, keyType=keyType) if len(valueMap) > 0: logging.info("Read %d items into map. EG: %s", len(valueMap), next(iter(valueMap.items()))) else: logging.warn("Read 0 items into value map!") else: valueMap = None # parse input files fileCounts = {} totals = {} fileLabels = {} sortedLabels = [] # Allow for file names to be preceded with TAG= for filename in arguments.input_files: bits = filename.split("=", 1) if len(bits) > 1: (filetag, filename) = bits else: filetag = filename fileLabels[filename] = filetag # keep order so that column order matches arguments sortedLabels.append(filetag) fileCounts[filetag] = {} totals[filetag] = 0 params = FilterParams.create_from_arguments(arguments) # TODO: incorporate weights into tophit algorithm! if arguments.countMethod == 'tophit': # Process all files at once and use overall abundance to pick best hits from edl import redistribute multifile = redistribute.multipleFileWrapper(fileLabels.items()) # don't give any hit translation, just use hit ids for redistribution readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, winnerTakeAll=True, parseStyle=arguments.parseStyle, sequenceWeights=sequenceWeights) # define method to turn Hits into Genes (kos, families) hitTranslator = getHitTranslator(parseStyle=arguments.parseStyle, hitStringMap=valueMap) # translateHit = lambda hit: hitTranslator.translateHit(hit)[0] # use read->file mapping and hit translator to get file based counts # from returned (read,Hit) pairs increment = 1 for (read_name, hit) in readHits: file_tag, read_name = read_name.split("/", 1) file_tag = unquote_plus(file_tag) gene = hitTranslator.translateHit(hit)[0] if gene is None: gene = "None" logging.debug( "READ: %s\t%s\t%s\t%s", file_tag, read_name, hit.hit, gene) genecount = fileCounts[file_tag].setdefault(gene, 0) if sequenceWeights is not None: increment = sequenceWeights.get(read_name, 1) fileCounts[file_tag][gene] = genecount + increment totals[file_tag] += increment logging.debug(str(totals)) else: # Original way, just process each file separately for (filename, filetag) in fileLabels.items(): infile = open(filename, 'rU') hitIter = parseM8FileIter(infile, valueMap, params, arguments.parseStyle, arguments.countMethod, ignoreEmptyHits=arguments.mappedHitsOnly) (total, counts, hitMap) = \ countIterHits(hitIter, allMethod=arguments.allMethod, weights=sequenceWeights) fileCounts[filetag] = counts totals[filetag] = total logging.info( "parsed %d hits (%d unique) for %d reads from %s", total, len(counts), len(hitMap), filename) infile.close() logging.debug(repr(fileCounts)) printCountTablesByLevel(fileCounts, totals, sortedLabels, arguments)
def main(): description = __doc__ parser = argparse.ArgumentParser(description=description) parser.add_argument("input_files", nargs="+", default=[], metavar="INFILE", help="List of hit tables to process") parser.add_argument("-o", "--outfile", dest="output_file", metavar="OUTFILE", help="Write count table to OUTFILE") parser.add_argument("-l", "--level", dest="levels", default=None, metavar="LEVEL", action="append", help=""" Level(s) to collect counts on. Use flag multiple times to specify multiple levels. If multiple values given, one table produced for each with rank name appended to file name. Levels can be an integer (1-3) for KEGG or SEED levels, any one of 'gene', 'role', 'family', 'ko', or 'ortholog' (which are all synonyms), or anything not synonymous with 'gene' to get CAZy groups. Defaults to ortholog/role and levels 1, 2, and 3 for KEGG and SEED and gene and group for CAZy and COG.""") # option for deconvoluting clusters or assemblies add_weight_arguments(parser, multiple=True) # cutoff options add_count_arguments(parser) # format, ortholog heirarchy, and more kegg.add_path_arguments( parser, defaults={'countMethod': 'tophit'}, choices={ 'countMethod': ('tophit', 'first', 'most', 'all', 'consensus') }, helps={ 'countMethod': ("How to deal with counts from multiple hits. ('first': " "just use the first hit, 'most': " "can return multiple hits, 'all': return every hit, " "consensus: return None unless all the same). Do not " "use most or consensus with more than one level at a time. " "Default is 'tophit': This breaks any ties by choosing " "the most abundant hit based on other unambiguous " "assignments.") }) # log level and help add_universal_arguments(parser) arguments = parser.parse_args() setup_logging(arguments) if len(arguments.input_files) == 0: parser.error("Must supply at least one m8 file to parse") # Set defaults and check for some conflicts if arguments.levels is None and arguments.heirarchyFile is None: # using hit names only arguments.levels = [None] else: if arguments.heirarchyFile is None \ and arguments.heirarchyType != 'cazy': logging.warning("Type: %s", arguments.heirarchyType) parser.error("Cannot select levels without a heirarchy (ko) file") if arguments.levels is None: # set a default if arguments.heirarchyType is 'kegg': arguments.levels = ['ko', '1', '2', 'pathway'] if arguments.heirarchyType is 'seed': arguments.levels = ['role', '1', '2', 'subsystem'] else: arguments.levels = ['gene', 'group'] try: # Make sure the rank lists make sense arguments.levels = cleanLevels(arguments.levels) except Exception as e: parser.error(str(e)) # load weights file sequenceWeights = loadSequenceWeights(arguments.weights) # only print to stdout if there is a single level if len(arguments.levels) > 1 and arguments.output_file is None: parser.error("STDOUT only works if a single level is chosen!") cutoff = arguments.cutoff # map reads to hits if arguments.mapFile is not None: if arguments.mapStyle == 'auto': with open(arguments.mapFile) as f: firstLine = next(f) while len(firstLine) == 0 or firstLine[0] == '#': firstLine = next(f) if koMapRE.search(firstLine): arguments.mapStyle = 'kegg' elif seedMapRE.search(firstLine): arguments.mapStyle = 'seed' elif tabMapRE.search(firstLine): arguments.mapStyle = 'tab' # elif cogMapRE.search(firstLine): # arguments.mapStyle='cog' else: raise Exception( "Cannot figure out map type from first line:\n%s" % (firstLine)) logging.info("Map file seems to be: %s", arguments.mapStyle) if arguments.mapStyle == 'kegg': valueMap = kegg.parseLinkFile(arguments.mapFile) elif arguments.mapStyle == 'seed': valueMap = kegg.parseSeedMap(arguments.mapFile) # elif arguments.mapStyle=='cog': # valueMap=kegg.parseCogMap(arguments.mapFile) else: if arguments.parseStyle == GIS: keyType = int else: keyType = None valueMap = parseMapFile(arguments.mapFile, valueType=None, valueDelim=arguments.tab_map_delim, keyType=keyType) if len(valueMap) > 0: logging.info("Read %d items into map. EG: %s", len(valueMap), next(iter(valueMap.items()))) else: logging.warn("Read 0 items into value map!") else: valueMap = None # parse input files fileCounts = {} totals = {} fileLabels = {} sortedLabels = [] # Allow for file names to be preceded with TAG= for filename in arguments.input_files: bits = filename.split("=", 1) if len(bits) > 1: (filetag, filename) = bits else: filetag = filename fileLabels[filename] = filetag # keep order so that column order matches arguments sortedLabels.append(filetag) fileCounts[filetag] = {} totals[filetag] = 0 # TODO: incorporate weights into tophit algorithm! if arguments.countMethod == 'tophit': # Process all files at once and use overall abundance to pick best hits from edl import redistribute params = FilterParams.create_from_arguments(arguments) multifile = redistribute.multipleFileWrapper(fileLabels.items()) # don't give any hit translation, just use hit ids for redistribution readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, winnerTakeAll=True, parseStyle=arguments.parseStyle, sequenceWeights=sequenceWeights) # define method to turn Hits into Genes (kos, families) hitTranslator = getHitTranslator(parseStyle=arguments.parseStyle, hitStringMap=valueMap) # translateHit = lambda hit: hitTranslator.translateHit(hit)[0] # use read->file mapping and hit translator to get file based counts # from returned (read,Hit) pairs increment = 1 for (read_name, hit) in readHits: file_tag, read_name = read_name.split("/", 1) file_tag = unquote_plus(file_tag) gene = hitTranslator.translateHit(hit)[0] if gene is None: gene = "None" logging.debug("READ: %s\t%s\t%s\t%s", file_tag, read_name, hit.hit, gene) genecount = fileCounts[file_tag].setdefault(gene, 0) if sequenceWeights is not None: increment = sequenceWeights.get(read_name, 1) fileCounts[file_tag][gene] = genecount + increment totals[file_tag] += increment logging.debug(str(totals)) else: # Original way, just process each file separately for (filename, filetag) in fileLabels.items(): infile = open(filename, 'rU') hitIter = parseM8FileIter(infile, valueMap, arguments.hitTableFormat, arguments.filter_top_pct, arguments.parseStyle, arguments.countMethod, ignoreEmptyHits=arguments.mappedHitsOnly) (total, counts, hitMap) = \ countIterHits(hitIter, allMethod=arguments.allMethod, weights=sequenceWeights) fileCounts[filetag] = counts totals[filetag] = total logging.info("parsed %d hits (%d unique) for %d reads from %s", total, len(counts), len(hitMap), filename) infile.close() logging.debug(repr(fileCounts)) printCountTablesByLevel(fileCounts, totals, sortedLabels, arguments)
def main(): """" Set up the CLI """ parser = argparse.ArgumentParser(description=__doc__) parser.add_argument("input_files", nargs="+", default=[], metavar="INFILE", help="List of hit tables to process") parser.add_argument("-o", "--outfile", dest="outfile", metavar="OUTFILE", help="Write count table to OUTFILE") parser.add_argument("-r", "--rank", dest="ranks", default=None, metavar="RANK", action="append", help=""" Rank(s) to collect counts on. Use flag multiple times to specify multiple ranks. If multiple values given, one table produced for each with rank name appended to file name. Defaults to all major ranks between phylum and species. Corresponds to rank names in nodes.dmp. To see list run: 'cut -f5 nodes.dmp | uniq | sort | uniq' in ncbi tax dir. Will also accept 'organism' to mean no rank (ie, just the organism name).""") parser.add_argument( "-s", "--collapseToDomain", default=False, action="store_true", help="Collapse all taxa below given rank down to " "superkingdom/domain. EG: in the genus output, anything " "assigned to Cyanobactia, will be lumped in with all " "other bacteria") parser.add_argument( "-R", "--printRank", dest="printRanks", action="append", help="Include indeicated rank(s) in lineage of printed taxa. " "Will be ignored if beyond the rank of the taxa " "(IE We can't include species if the taxon being counted " "is genus)") # option for deconvoluting clusters or assemblies add_weight_arguments(parser, multiple=True) # cutoff options add_count_arguments(parser) # format, tax dir, and more add_taxon_arguments(parser, choices={ 'countMethod': ('LCA', 'all', 'first', 'most', 'tophit', 'toporg', 'consensus') }) # log level and help add_universal_arguments(parser) arguments = parser.parse_args() setup_logging(arguments) if len(arguments.input_files) == 0: parser.error("Must supply at least one m8 file to parse") # Handle the case where Galaxy tries to set None as a string arguments.ranks = checkNoneOption(arguments.ranks) arguments.printRanks = checkNoneOption(arguments.printRanks) logging.info("Printing out ranks: %r", arguments.ranks) # Set defaults and check for some conflicts if arguments.ranks is None and arguments.taxdir is None: # using hit names only arguments.ranks = [ORG_RANK] if arguments.printRanks is not None: parser.error("Display ranks are not used without taxonomic info") else: if arguments.taxdir is None: parser.error("Cannot select ranks without a taxonomy") if arguments.ranks is None: # set a default arguments.ranks = [ 'phylum', 'class', 'order', 'family', 'genus', 'species' ] try: # Make sure the rank lists make sense arguments.ranks = cleanRanks(arguments.ranks) if arguments.printRanks is not None: arguments.printRanks = cleanRanks(arguments.printRanks) except Exception as e: parser.error(str(e)) # load weights file sequenceWeights = loadSequenceWeights(arguments.weights) # only print to stdout if there is a single rank if len(arguments.ranks) > 1 and arguments.outfile is None: parser.error("STDOUT only works if a single rank is chosen!") # Because rank is used in parsing hits, we can only do multiple ranks for # certain kinds of count methods if len(arguments.ranks) > 1: rank = None if arguments.countMethod in ['consensus', 'most']: parser.error( "Using multiple ranks does not work with the 'consensus' " "or 'most' counting methods. LCA should give the same " "results as consensus. If you really want to do this, " "use a bash loop:'for rank in phylum order genus; do " "COMMAND -r ${rank}; done'") else: rank = arguments.ranks[0] # load necessary maps (taxonomy, hitStringMap) = readMaps(arguments) # parse input files fileCounts = {} totals = {} fileLabels = {} sortedLabels = [] # Allow for file names to be preceded with TAG= for filename in arguments.input_files: bits = filename.split("=", 1) if len(bits) > 1: (filetag, filename) = bits else: filetag = filename fileLabels[filename] = filetag # keep order so that column order matches arguments sortedLabels.append(filetag) fileCounts[filetag] = {} totals[filetag] = 0 if arguments.countMethod == 'tophit' or arguments.countMethod == 'toporg': # Process all files at once and use overall abundance to pick best hits from edl import redistribute params = FilterParams.create_from_arguments(arguments) multifile = redistribute.multipleFileWrapper(fileLabels.keys()) if arguments.countMethod == 'tophit': # don't give any taxonomy, just map to accessions for # redistribution readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, winnerTakeAll=True, parseStyle=arguments.parseStyle, sequenceWeights=sequenceWeights) # define method to turn Hits into orgnaisms hitTranslator = getHitTranslator(parseStyle=arguments.parseStyle, taxonomy=taxonomy, hitStringMap=hitStringMap) translateHit = lambda hit: hitTranslator.translateHit(hit=hit)[0] else: # translate to organism before finding most abundant readHits = redistribute.pickBestHitByAbundance( multifile, filterParams=params, returnLines=False, returnTranslations=True, winnerTakeAll=True, taxonomy=taxonomy, hitStringMap=hitStringMap, parseStyle=ACCS) # Organisms will be returned, make translator trivial: translateHit = passThrough # use read->file mapping and hit translator to get file based counts # from returned (read,Hit) pairs increment = 1 for (read_name, hit) in readHits: file_name, read_name = read_name.split("/", 1) file_tag = fileLabels[unquote_plus(file_name)] taxon = translateHit(hit) taxcount = fileCounts[file_tag].setdefault(taxon, 0) if sequenceWeights is not None: increment = sequenceWeights.get(read_name, 1) fileCounts[file_tag][taxon] = taxcount + increment totals[file_tag] += increment logging.debug(str(totals)) else: # Original way, just process each file separately for (filename, filetag) in fileLabels.items(): infile = open(filename, 'rU') hitIter = parseM8FileIter(infile, hitStringMap, arguments.hitTableFormat, arguments.filter_top_pct, arguments.parseStyle, arguments.countMethod, taxonomy=taxonomy, rank=rank) (total, counts, hitMap) = \ countIterHits(hitIter, allMethod=arguments.allMethod, weights=sequenceWeights) fileCounts[filetag] = counts totals[filetag] = total logging.info("parsed %d hits (%d unique) for %d reads from %s", total, len(counts), len(hitMap), filename) infile.close() printCountTablesByRank(fileCounts, totals, sortedLabels, arguments)