def make_kfold_datasets(k, fastafile, firefile, outpre):
logging.warning("Making %s fold datasets" % k)
# shuffle all the fasta entries in place
np.random.shuffle(fastafile.names)
# create k folds out of the names as a list of lists
size = len(fastafile.names)
folds = [firefile.names[i::k] for i in range(k)]
# loop through each fold
for test_idx in range(len(folds)):
# make a seperate fasta file and fire file for the test fold
this_test_fasta = fa.FastaFile()
this_test_fire = FIREfile()
for name in folds[test_idx]:
this_test_fasta.add_entry(fastafile.pull_entry(name))
this_test_fire.add_entry(name, firefile.pull_value(name))
with open(outpre + "_test_%i.fa" % test_idx, mode="w") as outf:
this_test_fasta.write(outf)
this_test_fire.write(outpre + "_test_%i.txt" % test_idx)
# make a sperate fasta and fire file for the train folds
this_train_fasta = fa.FastaFile()
this_train_fire = FIREfile()
for train_idx in range(len(folds)):
if train_idx != test_idx:
for name in folds[train_idx]:
this_train_fasta.add_entry(fastafile.pull_entry(name))
this_train_fire.add_entry(name, firefile.pull_value(name))
logging.warning("Writing fold %i" % test_idx)
with open(outpre + "_train_%i.fa" % test_idx, mode="w") as outf:
this_train_fasta.write(outf)
this_train_fire.write(outpre + "_train_%i.txt" % test_idx)
def out_fasta(gb, features = ["CDS"], qual_names = ["locus_tag"]):
import fasta as fa
out_fasta = fa.FastaFile()
fa_name = set()
for feature in gb.features:
if feature.type in features:
# skip features with partial ends
if check_partial_start(feature.location.start) or check_partial_end(feature.location.end):
continue
# often times things need a gene name, thus will replace with locus tag or
# if that doesn't exist. NA_number
unique_name = check_qualifiers(feature, qual_names)
num = 1
id_name = unique_name
while id_name in fa_name:
id_name = unique_name + "_%d"%(num)
num += 1
fa_name.add(id_name)
header = ">" + id_name + \
" " + \
str(feature.qualifiers.get("product", ["NA"])[0])
seq = str(feature.extract(gb.seq))
entry = fa.FastaEntry(header = header, seq = seq)
out_fasta.add_entry(entry)
out_fasta.write(sys.stdout)
def read_genome(infile):
# sys.path.append("/home/mbwolfe/src/circ_mapper")
import fasta
genome = fasta.FastaFile()
with open(infile, mode="r") as f:
genome.read_whole_file(f)
chrm = genome.pull_entry(genome.chrm_names()[0])
return chrm
def out_fna(gb, chrm=None):
import fasta as fa
out_fasta = fa.FastaFile()
if chrm:
header = ">" + str(chrm)
else:
header = ">" + str(gb.name)
seq = str(gb.seq)
entry = fa.FastaEntry(header=header, seq=seq)
out_fasta.add_entry(entry)
out_fasta.write(sys.stdout)
def seq_file (file, format=None, revcomp=False, name="", gap=None, contig=None):
if (format == None): format = infer_format(file)
if (contig != None) and (format not in ["fasta",None]):
raise ValueError("Contigs are not supported for format %s" % format)
if (format == "fasta"): return fasta.FastaFile (file, revcomp=revcomp, name=name, gap=gap, contig=contig)
elif (format == "nib"): return nib.NibFile (file, revcomp=revcomp, name=name, gap=gap)
elif (format == "qdna"): return qdna.QdnaFile (file, revcomp=revcomp, name=name, gap=gap)
else:
if (format == None): format = ""
else: format = " " + format
raise ValueError("Unknown sequence format%s in %s" % (format,file.name))
def out_fasta(gb, features=["CDS"], qual_name="locus_tag"):
import fasta as fa
out_fasta = fa.FastaFile()
fa_name = set()
for NA_num, feature in enumerate(gb.features):
if feature.type in features:
# often times things need a gene name, thus will replace with locus tag or
# if that doesn't exist. NA_number
unique_name = str(
feature.qualifiers.get(qual_name,
["%s_%d" % ("NA", NA_num)])[0])
num = 1
id_name = unique_name
while id_name in fa_name:
id_name = unique_name + "_%d" % (num)
num += 1
fa_name.add(id_name)
header = ">" + id_name + \
" " + \
str(feature.qualifiers.get("product", ["NA"])[0])
seq = str(feature.extract(gb.seq))
entry = fa.FastaEntry(header=header, seq=seq)
out_fasta.add_entry(entry)
out_fasta.write(sys.stdout)
def dense_sampling_main(args):
# parse arguments
args = parser.parse_args()
# figure out random seed
np.random.seed(args.seed)
# read in genome
genome = fa.FastaFile()
logging.warning("reading in full genome")
with open(args.fasta) as inf:
genome.read_whole_file(inf)
# read in bed file
inbed = bed.BedFile()
logging.warning("reading in bed")
inbed.from_bed_file(args.bedfile)
# discretize the genome by size of window and step of window
outbed = discretize_genome(args.wsize, args.stepsize, genome)
# convert input bed to an interval file by chromosome
intervals = {chrm.chrm_name(): it.Intervals() for chrm in genome}
for feature in inbed:
this_interval = intervals[feature["chrm"]]
this_interval.add_interval(
it.Interval(feature["start"], feature["end"]))
intervals[feature["chrm"]] = this_interval
# figure out which intervals overlap and which don't
logging.warning("determining which intervals overlap")
positive_bed = bed.BedFile()
negative_bed = bed.BedFile()
for i, window in enumerate(outbed):
if i % 10000 == 0:
logging.warning("Checking interval %s" % i)
this_chrm = window["chrm"]
this_intervals = intervals[this_chrm]
window_interval = it.Interval(window["start"], window["end"])
perc_overlap = this_intervals.check_percent_overlap(window_interval)
if perc_overlap >= args.perc_overlap:
positive_bed.add_entry(window)
else:
negative_bed.add_entry(window)
# make fire file
fire = FIREfile()
out_fasta = fa.FastaFile()
for feature in positive_bed:
this_chrm = feature["chrm"]
this_name = feature["name"]
this_start = feature["start"]
this_end = feature["end"]
fire.add_entry(this_name, args.default_score)
out_fasta.add_entry(
fa.FastaEntry(
">" + this_name,
genome.pull_entry(this_chrm).pull_seq(this_start, this_end)))
for feature in negative_bed:
this_chrm = feature["chrm"]
this_name = feature["name"]
this_start = feature["start"]
this_end = feature["end"]
fire.add_entry(this_name, args.rand_score)
out_fasta.add_entry(
fa.FastaEntry(
">" + this_name,
genome.pull_entry(this_chrm).pull_seq(this_start, this_end)))
# write files
if args.true_bed:
positive_bed.write_bed_file(args.outpre + "_true.bed")
if args.rand_bed:
negative_bed.write_bed_file(args.outpre + "_rand.bed")
if not args.no_fasta:
with open(args.outpre + ".fa") as outf:
out_fasta.write(outf)
fire.write(args.outpre + "_fire.txt")
default=25,
help="Step size for dense mode, default = 25")
parser.add_argument('--perc_overlap',
type=float,
default=0.5,
help="Overlap cutoff for dense mode, default = 0.5")
# parse arguments
args = parser.parse_args()
if args.dense:
dense_sampling_main(args)
sys.exit()
# figure out random seed
np.random.seed(args.seed)
# read in genome
genome = fa.FastaFile()
logging.warning("reading in full genome")
with open(args.fasta) as inf:
genome.read_whole_file(inf)
# read in bed file
inbed = bed.BedFile()
logging.warning("reading in bed")
inbed.from_bed_file(args.bedfile)
# check how much of the genome the regions cover
genome_length = {}
total_length = 0
for chrm in genome:
genome_length[chrm] = len(chrm)
total_length += len(chrm)
import argparse
parser = argparse.ArgumentParser("Combine fasta files into one fasta file")
parser.add_argument('outfilepre', type=str, help="prefix for output files")
parser.add_argument('infiles',
type=str,
nargs='+',
help="files to combine")
parser.add_argument(
'--masked_regions',
type=str,
help="A .bed file containing a list of regions to replace with Ns")
args = parser.parse_args()
genome_name = args.outfilepre
fasta_files = args.infiles
all_fastas = [fa.FastaFile() for fafile in fasta_files]
# read in all fastas
for fafile, fafilename in zip(all_fastas, fasta_files):
with open(fafilename, mode="r") as inf:
fafile.read_whole_file(inf)
final_fasta = fa.FastaFile()
for fafile in all_fastas:
for entry in fafile:
final_fasta.add_entry(entry)
lengths = {}
for entry in final_fasta:
chrm_name = entry.chrm_name()
lengths[chrm_name] = len(entry)
# Replace masked regions with Ns