def get_acceptable_species():
"""
Finds all species in data directory
:return: acceptable_species (set): string of available genome
"""
acceptable_species = set([])
for fn in os.listdir(clipper.data_dir()):
fn = fn.split(".")[0]
if fn == "__init__":
continue
acceptable_species.add(fn)
return acceptable_species
def make_features(self):
Region = collections.namedtuple("Region", ["region", "gene_id"])
bedtracks = {}
for region in self.assigned_regions:
bedtracks[region] = pybedtools.BedTool(
os.path.join(clipper.data_dir(), "regions", "%s_%s.bed" % (self.species, region))
)
features = HTSeq.GenomicArrayOfSets("auto", stranded=True)
for region, bedtrack in bedtracks.items():
for iv, interval in izip(CLIP_analysis.bed_to_genomic_interval(bedtrack), bedtrack):
features[iv] = set([Region(region, interval.name)])
return features
def count_genomic_region_sizes(regions, species="hg19"):
"""
Counts the genomic region sizes for the specified regions dir
(should be changed to GTF)
"""
genomic_region_sizes = {}
#TODO update this to work of GFF file, because something isn't matching up...
for region in regions:
region_tool = pybedtools.BedTool(os.path.join(clipper.data_dir(), "regions", species + "_" + region + ".bed"))
genomic_region_sizes[region] = region_tool.total_coverage()
return genomic_region_sizes
def make_features(self):
Region = collections.namedtuple("Region", ["region", "gene_id"])
bedtracks = {}
for region in self.assigned_regions:
bedtracks[region] = pybedtools.BedTool(
os.path.join(clipper.data_dir(), "regions",
"%s_%s.bed" % (self.species, region)))
features = HTSeq.GenomicArrayOfSets("auto", stranded=True)
for region, bedtrack in bedtracks.items():
for iv, interval in izip(
CLIP_analysis.bed_to_genomic_interval(bedtrack), bedtrack):
features[iv] = set([Region(region, interval.name)])
return features
def get_acceptable_species():
"""
Finds all species in data directory
"""
acceptable_species = set([])
for fn in os.listdir(clipper.data_dir()):
fn = fn.split(".")[0]
if fn == "__init__":
continue
acceptable_species.add(fn)
return acceptable_species
def __init__(self, species, db=None, regions_dir=None):
"""
creates genomic features function, chooses
how to direct creation of features based on _species
regions_dir : str location to create region
species: str species (hg19, mm9, ce10
db: gffutils FeatureDb object
"""
if regions_dir == None:
regions_dir = os.path.join(data_dir(), "regions")
self._regions_dir = regions_dir
self._db = db
self._species = species
if species in ["hg19", "mm9"]:
self._feature_names = {
"five_prime_utr": "five_prime_utr",
"three_prime_utr": "three_prime_utr",
"exon": "exon",
"CDS": "CDS",
"gene_id": "gene_id",
"transcript": "transcript",
}
self._fix_chrom = self._fix_chrom_null
elif species in ["ce10"]:
self._feature_names = {
"five_prime_utr": "five_prime_UTR",
"three_prime_utr": "three_prime_UTR",
"exon": "exon",
"CDS": "CDS",
"gene_id": "ID",
"transcript": "mRNA",
}
self._fix_chrom = self._fix_chrom_ce10
if self._db is not None:
self.featuretypes = list(self._db.featuretypes())
else:
self.featuretypes = None
def __init__(self, species, db=None, regions_dir=None):
"""
creates genomic features function, chooses
how to direct creation of features based on _species
regions_dir : str location to create region
species: str species (hg19, mm9, ce10
db: gffutils FeatureDb object
"""
if regions_dir == None:
regions_dir = os.path.join(data_dir(), "regions")
self._regions_dir = regions_dir
self._db = db
self._species = species
if species in ["hg19", "mm9"]:
self._feature_names = {
"five_prime_utr" : "five_prime_utr",
"three_prime_utr" : "three_prime_utr",
"exon" : "exon",
"CDS" : "CDS",
"gene_id" : "gene_id",
"transcript" : "transcript",
}
self._fix_chrom = self._fix_chrom_null
elif species in ["ce10"]:
self._feature_names = {
"five_prime_utr" : "five_prime_UTR",
"three_prime_utr" : "three_prime_UTR",
"exon" : "exon",
"CDS" : "CDS",
"gene_id" : "ID",
"transcript" : "mRNA",
}
self._fix_chrom = self._fix_chrom_ce10
if self._db is not None:
self.featuretypes = list(self._db.featuretypes())
else:
self.featuretypes = None
def get_uORF_start_stop_gff(self):
"""
Returns hg19 uORFs
"""
db = gffutils.FeatureDB(
"/nas3/yeolab/Genome/ensembl/gtf/gencode.v17.annotation.gtf.db.old"
)
transcript_gene_dict = self._create_transcript_map(db)
#get all 5' UTRs
(UTR3, UTR5, exons, genes, introns,
CDS) = CLIP_analysis.get_genomic_regions(
os.path.join(clipper.data_dir(), "regions"), "hg19", db).values()
five_prime_utr_dict = self._get_five_prime_utr_sequences(
UTR5, "/nas3/yeolab/Genome/ucsc/hg19/chromosomes/all.fa")
return self._get_uorf_start_stop(five_prime_utr_dict)
def get_uORF_start_stop_gff(self):
"""
Returns hg19 uORFs
"""
db = gffutils.FeatureDB("/nas3/yeolab/Genome/ensembl/gtf/gencode.v17.annotation.gtf.db.old")
transcript_gene_dict = self._create_transcript_map(db)
#get all 5' UTRs
(UTR3, UTR5,
exons, genes,
introns, CDS) = CLIP_analysis.get_genomic_regions(os.path.join(clipper.data_dir(), "regions"),
"hg19",
db).values()
five_prime_utr_dict = self._get_five_prime_utr_sequences(UTR5, "/nas3/yeolab/Genome/ucsc/hg19/chromosomes/all.fa")
return self._get_uorf_start_stop(five_prime_utr_dict)
def get_exon_bed(species):
short_species = species.split("_")[0]
return os.path.join(clipper.data_dir(), "regions",
"%s_%s.bed" % (short_species, "exons"))
def __init__(self, species, db=None, regions_dir=None):
"""
creates genomic features function, chooses
how to direct creation of features based on _species
regions_dir : str location to create region
species: str species (hg19, mm9, ce10
db: gffutils FeatureDb object
"""
if regions_dir == None:
regions_dir = os.path.join(data_dir(), "regions")
self._regions_dir = regions_dir
self._db = db
self._species = species
# I'm going to be lazy and leave this here, its needed to make a new genomic features for human genomes
# engineering so it doesn't take too much time on load will be slightly annoying so just uncomment when you need it
# result = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))
# for x, feature in enumerate(db.all_features()):
# gene_ids = feature.attributes['gene_id']
# transcript_ids = feature.attributes['transcript_id']
# feature_type = feature.featuretype
#
#
# if feature_type == "gene":
# if len(gene_ids) != 1:
# print gene_ids[0]
# break
#
# result[gene_ids[0]]['gene'] = feature
# else:
# for gene_id in gene_ids:
# for transcript_id in transcript_ids:
# result[gene_id][transcript_id][feature_type].append(feature)
#
# self._feature_hash = result
if species in ["hg19", "mm9", "hg19_v19", "GRCh38_v24"]:
self._feature_names = {
"five_prime_utr": "five_prime_utr",
"three_prime_utr": "three_prime_utr",
"exon": "exon",
"CDS": "CDS",
"gene_id": "gene_id",
"transcript": "transcript",
}
self._fix_chrom = self._fix_chrom_null
elif species in ["ce10"]:
self._feature_names = {
"five_prime_utr": "five_prime_UTR",
"three_prime_utr": "three_prime_UTR",
"exon": "exon",
"CDS": "CDS",
"gene_id": "ID",
"transcript": "mRNA",
}
self._fix_chrom = self._fix_chrom_ce10
if self._db is not None:
self.featuretypes = list(self._db.featuretypes())
else:
self.featuretypes = None
def assign_to_regions(tool,
clusters=None,
assigned_dir=".",
species="hg19",
nrand=3):
"""
Assigns each cluster to a genic region
finally saves all generated bed and fasta files for future analysis...
tool - a bed tool (each line represnting a cluster)
clusters - name of cluster file (optional)
assigned_dir - location to save files in
species - str species to segment
nrand - int number offsets times to shuffle for null hypothesis
"""
if clusters is None:
clusters, ext = os.path.splitext(os.path.basename(tool.fn))
bedtracks = {}
regions, assigned_regions = regions_generator()
short_species = species.split("_")[0]
if short_species == "GRCh38":
short_species = "hg38"
for region in regions:
bedtracks[region] = pybedtools.BedTool(
os.path.join(clipper.data_dir(), "regions",
"%s_%s.bed" % (species, region)))
#creates the basics of bed dict
bed_dict = {'all': {'rand': {}}}
genes = pybedtools.BedTool(
os.path.join(clipper.data_dir(), "regions",
"%s_genes.bed" % (species)))
offsets = get_offsets_bed12(tool)
if tool.field_count() <= 5:
tool.sort().merge().saveas()
elif 6 <= tool.field_count() < 8:
#Hack to get around not having gene name assigned by peak caller, due to overlapping genes this won't be perfect
#move_name_real = functools.partial(move_name, original_length=len(tool[0].fields))
#tool = tool.intersect(genes, wo=True, s=True).each(move_name_real).saveas()
#fix_strand_ok = functools.partial(fix_strand, warn=False)
tool = tool.sort().merge(
s=True, c="4,5,6",
o="collapse,collapse,collapse").each(fix_strand_v26).saveas()
#elif not tool[0][7].isdigit():
# tool = tool.sort().merge(s=True, c="4,5,6", o="collapse,collapse,collapse").each(fix_strand).each(fix_name).saveas()
else: #Clipper, this is ideal we like this technique
tool = tool.sort().merge(s=True,
c="4,5,6,7,8",
o="collapse,collapse,collapse,min,min").each(
fix_strand_v26).saveas()
remaining_clusters = adjust_offsets(tool, offsets)
# print "There are a total %d clusters I'll examine" % (len(tool))
for region in regions:
remaining_clusters, overlapping = intersection(remaining_clusters,
b=bedtracks[region])
#if for some reason there isn't a peak in the region skip it
if len(overlapping) == 0:
# print "ignoring %s " % region
continue
#sets up bed dict for this region
bed_dict[region] = {
'real': overlapping.sort(stream=True).saveas(),
'rand': {}
}
no_overlapping_count = len(remaining_clusters)
overlapping_count = len(bed_dict[region]['real'])
# print "For region: %s found %d that overlap and %d that don't" % (region,
# overlapping_count,
# no_overlapping_count)
if 'real' not in bed_dict['all']:
bed_dict['all']['real'] = bed_dict[region]['real']
else:
bed_dict['all']['real'] = bed_dict['all']['real'].cat(
bed_dict[region]['real'], stream=True,
postmerge=False).saveas()
#saves offsets so after shuffling the offsets can be readjusted
offset_dict = get_offsets_bed12(bed_dict[region]['real'])
for i in range(nrand):
random_intervals = bed_dict[region]['real'].shuffle(
genome=short_species, incl=bedtracks[region].fn).sort()
random_intervals = fix_shuffled_strand(random_intervals,
bedtracks[region].fn)
random_intervals = adjust_offsets(random_intervals, offset_dict)
bed_dict[region]['rand'][i] = random_intervals.saveas()
if i not in bed_dict['all']['rand']:
bed_dict['all']['rand'][i] = bed_dict[region]['rand'][i]
else:
bed_dict['all']['rand'][i] = bed_dict['all']['rand'][i].cat(
bed_dict[region]['rand'][i], stream=True, postmerge=False)
#if there are no more clusters to assign stop trying
if no_overlapping_count == 0:
break
# print "After assigning %d un-categorized regions" % len(remaining_clusters)
if len(remaining_clusters) > 0:
bed_dict['uncatagorized'] = {
'real': remaining_clusters.sort(stream=True).saveas()
}
bed_dict = save_bedtools(bed_dict, clusters, assigned_dir)
return bed_dict
def __init__(self, species, db=None, regions_dir=None, gencode=False):
"""
creates genomic features function, chooses
how to direct creation of features based on _species
regions_dir : str location to create region
species: str species (hg19, mm9, ce10
db: gffutils FeatureDb object
"""
if regions_dir == None:
regions_dir = os.path.join(data_dir(), "regions")
self._regions_dir = regions_dir
self._db = db
self._species = species
#I'm going to be lazy and leave this here, its needed to make a new genomic features for human genomes
#engineering so it doesn't take too much time on load will be slightly annoying so just uncomment when you need it
result = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))
for x, feature in enumerate(db.all_features()):
gene_ids = feature.attributes['gene_id']
transcript_ids = feature.attributes['transcript_id']
feature_type = feature.featuretype
if feature_type == "gene":
if len(gene_ids) != 1:
print gene_ids[0]
break
result[gene_ids[0]]['gene'] = feature
else:
for gene_id in gene_ids:
for transcript_id in transcript_ids:
result[gene_id][transcript_id][feature_type].append(
feature)
self._feature_hash = result
if species in ["hg19", "mm9", "hg19_v19", "GRCh38_v24", "hb27"
] or gencode:
self._feature_names = {
"five_prime_utr": "five_prime_utr",
"three_prime_utr": "three_prime_utr",
"exon": "exon",
"CDS": "CDS",
"gene_id": "gene_id",
"transcript": "transcript",
}
self._fix_chrom = self._fix_chrom_null
elif species in ["ce10"]:
self._feature_names = {
"five_prime_utr": "five_prime_UTR",
"three_prime_utr": "three_prime_UTR",
"exon": "exon",
"CDS": "CDS",
"gene_id": "ID",
"transcript": "mRNA",
}
self._fix_chrom = self._fix_chrom_ce10
if self._db is not None:
self.featuretypes = list(self._db.featuretypes())
else:
self.featuretypes = None
def assign_to_regions(tool, clusters=None, assigned_dir=".", species="hg19", nrand=3):
"""
Assigns each cluster to a genic region
finally saves all generated bed and fasta files for future analysis...
tool - a bed tool (each line represnting a cluster)
clusters - name of cluster file (optional)
assigned_dir - location to save files in
species - str species to segment
nrand - int number offsets times to shuffle for null hypothesis
"""
if clusters is None:
clusters, ext = os.path.splitext(os.path.basename(tool.fn))
bedtracks = {}
regions, assigned_regions = regions_generator()
short_species = species.split("_")[0]
for region in regions:
bedtracks[region] = pybedtools.BedTool(os.path.join(clipper.data_dir(), "regions", "%s_%s.bed" % (species,
region)))
#creates the basics of bed dict
bed_dict = {'all': {'rand': {}}}
genes = pybedtools.BedTool(os.path.join(clipper.data_dir(), "regions", "%s_genes.bed" % (species)))
offsets = get_offsets_bed12(tool)
if tool.field_count() <= 5:
tool.sort().merge().saveas()
elif 6 <= tool.field_count() < 8:
#Hack to get around not having gene name assigned by peak caller, due to overlapping genes this won't be perfect
#move_name_real = functools.partial(move_name, original_length=len(tool[0].fields))
#tool = tool.intersect(genes, wo=True, s=True).each(move_name_real).saveas()
#fix_strand_ok = functools.partial(fix_strand, warn=False)
tool = tool.sort().merge(s=True, c="4,5,6", o="collapse,collapse,collapse").each(fix_strand).saveas()
elif not tool[0][7].isdigit():
tool = tool.sort().merge(s=True, c="4,5,6", o="collapse,collapse,collapse").each(fix_strand).saveas()
else: #Clipper, this is ideal we like this technique
tool = tool.sort().merge(s=True, c="4,5,6,7,8", o="collapse,collapse,collapse,min,min").each(fix_strand).saveas()
remaining_clusters = adjust_offsets(tool, offsets)
# print "There are a total %d clusters I'll examine" % (len(tool))
for region in regions:
remaining_clusters, overlapping = intersection(remaining_clusters, b=bedtracks[region])
#if for some reason there isn't a peak in the region skip it
if len(overlapping) == 0:
# print "ignoring %s " % region
continue
#sets up bed dict for this region
bed_dict[region] = {'real': overlapping.sort(stream=True).saveas(),
'rand': {}}
no_overlapping_count = len(remaining_clusters)
overlapping_count = len(bed_dict[region]['real'])
# print "For region: %s found %d that overlap and %d that don't" % (region,
# overlapping_count,
# no_overlapping_count)
if 'real' not in bed_dict['all']:
bed_dict['all']['real'] = bed_dict[region]['real']
else:
bed_dict['all']['real'] = bed_dict['all']['real'].cat(bed_dict[region]['real'], stream=True, postmerge=False).saveas()
#saves offsets so after shuffling the offsets can be readjusted
offset_dict = get_offsets_bed12(bed_dict[region]['real'])
for i in range(nrand):
random_intervals = bed_dict[region]['real'].shuffle(genome=short_species, incl=bedtracks[region].fn).sort()
random_intervals = fix_shuffled_strand(random_intervals, bedtracks[region].fn)
random_intervals = adjust_offsets(random_intervals, offset_dict)
bed_dict[region]['rand'][i] = random_intervals.saveas()
if i not in bed_dict['all']['rand']:
bed_dict['all']['rand'][i] = bed_dict[region]['rand'][i]
else:
bed_dict['all']['rand'][i] = bed_dict['all']['rand'][i].cat(bed_dict[region]['rand'][i], stream=True, postmerge=False)
#if there are no more clusters to assign stop trying
if no_overlapping_count == 0:
break
# print "After assigning %d un-categorized regions" % len(remaining_clusters)
if len(remaining_clusters) > 0:
bed_dict['uncatagorized'] = {'real': remaining_clusters.sort(stream=True).saveas()}
bed_dict = save_bedtools(bed_dict, clusters, assigned_dir)
return bed_dict
def get_exon_bed(species):
short_species = species.split("_")[0]
return os.path.join(clipper.data_dir(), "regions", "%s_%s.bed" % (short_species, "exons"))
