def fold_groups(seqs, struct, hold, numkept=None):
'''Function for multithreading.
Recomputes structure for a group'''
try:
aln, currstruct = bayesfold(seqs)
if currstruct == "":
currstruct = struct
hold[currstruct] = structgroups[struct]
except Exception, e:
print str(e)
stdout.flush()
def fold_clusters(lock, cluster, seqs, otufolder):
'''Function for multithreading.
Computes structure for a cluster and writes it to file'''
aln, struct = bayesfold(seqs, params={"-diags": True})
gshape = get_shape(struct)
#write structure out to file
lock.acquire()
cfo = open(otufolder + "cluster_structs.fasta", 'a')
cfo.write(">" + cluster + " " + gshape + "\n" + struct + "\n")
cfo.close()
#print cluster + ": " + struct
#stdout.flush()
lock.release()
def fold_clusters(lock, cluster, seqs, otufile):
'''Function for multithreading.
Computes structure for a cluster and writes it to file'''
aln, struct = bayesfold(seqs, params={"-diags": True})
# write structure out to file
try:
lock.acquire()
cfo = open(otufile, 'a')
cfo.write(">%s\n%s\n" % (cluster, struct))
cfo.write(aln.toFasta() + "\n")
cfo.close()
lock.release()
except Exception:
lock.release()
def run_fold_for_infernal(currgroup, groupfasta, basefolder, minseqs=1):
'''Function for multithreading. Creates the final BayesFold alignment and
writes to files, then r2r struct'''
try:
#run locana-p on the superclusters to get alignment and structure
#skip if already run and program just crashed or whatever
currotufolder = basefolder + "group_" + str(currgroup)
if exists(currotufolder):
return ""
seqs = []
count = 0
out = "group " + str(currgroup) + ": "
for header, seq in MinimalFastaParser(open(groupfasta, 'rU')):
seqs.append((header.split()[0] + "_" + header.split("_")[1], seq))
count += int(header.split("_")[1])
out += "\n" + str(count) + " sequences\n"
if count < minseqs:
return ""
stdout.flush()
#hard limit of 500 sequences to align and fold for memory reasons
if len(seqs) > 500:
seqs = seqs[:500]
#run BayesFold on sequences in the group
#maxiters set to 5 because should have huge amount of sequences for some groups
aln, struct = bayesfold(seqs, params={"-diags": True})
#create output folder for group
mkdir(currotufolder)
out += str(aln.getNumSeqs()) + " unique sequences\n"
out += "Structure: " + struct + "\n"
#write out alignment and structure in fasta and stockholm formats
#write that shit
logout = open(currotufolder + "/log.txt", 'w')
logout.write(out)
logout.close()
alnout = open(currotufolder + "/bayesfold-aln.fasta", 'w')
alnout.write(aln.toFasta() + "\n>SS_struct\n" + struct + "\n")
alnout.close()
alnout = open(currotufolder + "/bayesfold-aln.sto", 'w')
struct_dict = {'SS_cons': struct}
alnout.write(stockholm_from_alignment(aln, GC_annotation=struct_dict))
alnout.close()
#make R2R secondary structure for alignment
make_r2r(currotufolder + "/bayesfold-aln.sto", currotufolder, "group_" + str(currgroup))
except Exception, e:
print str(e)
stdout.flush()
def fold_clusters(lock, cluster, seqs, otufolder):
'''Function for multithreading.
Computes structure for a cluster and writes it to file'''
try:
aln, struct = bayesfold(seqs)
#write structure out to file
lock.acquire()
cfo = open(otufolder + "cluster_structs.fasta", 'a')
cfo.write(">" + cluster + "\n" + struct + "\n")
cfo.close()
lock.release()
#print cluster + ": " + struct
#stdout.flush()
except Exception, e:
cluster, struct, "\nERROR!"
stdout.flush()
lock.release()
def run_fold_for_infernal(currgroup, groupfasta, basefolder, minseqs=1):
'''Function for multithreading
creates the final BayesFold alignment and writes to files, then r2r struct'''
try:
#run locana-p on the superclusters to get the alignment and consensus structure
#skip if already run and program just crashsed or whatever
currotufolder = basefolder + "group_" + str(currgroup)
if exists(currotufolder):
return ""
seqs = []
count = 0
out = "group " + str(currgroup) + ": "
for header, seq in MinimalFastaParser(open(groupfasta, 'rU')):
seqs.append((header.split()[0] + "_" + header.split("_")[1], seq))
count += int(header.split("_")[1])
out += "\n" + str(count) + " sequences\n"
if count < minseqs:
return ""
#make sure group has enough sequences before continuing
#run BayesFold on the at most 50 most abundant sequences in the group
aln, struct = bayesfold(seqs)
#create output folder for group
mkdir(currotufolder)
out += str(aln.getNumSeqs()) + " unique sequences\n"
out += "Structure: " + struct + "\n"
#write out alignment and structure in fasta and stockholm formats
#write that shit
logout = open(currotufolder + "/log.txt", 'w')
logout.write(out)
logout.close()
alnout = open(currotufolder + "/bayesfold-aln.fasta", 'w')
alnout.write(aln.toFasta() + "\n>SS_struct\n" + struct + "\n")
alnout.close()
alnout = open(currotufolder + "/bayesfold-aln.sto", 'w')
struct_dict = {'SS_cons': struct}
alnout.write(stockholm_from_alignment(aln, GC_annotation=struct_dict))
alnout.close()
#make R2R secondary structure for alignment
make_r2r(currotufolder + "/bayesfold-aln.sto", currotufolder, "group_" + str(currgroup))
except Exception, e:
print str(e)
stdout.flush()
def fold_clusters(lock, cluster, seqs, otufolder):
'''Function for multithreading.
Computes structure for a cluster and writes it to file'''
#assuming that the fasta has 10 or more sequences in it. Safe assumption
#if this is a significant cluster
#only using 10 because this is initial structure calc so needs to be fast
#and this gives a very good approximation for the initial rnaforester grouping
try:
aln, struct = bayesfold(seqs)
#write structure out to file
lock.acquire()
cfo = open(otufolder + "cluster_structs.fasta", 'a')
cfo.write(">" + cluster + "\n" + struct + "\n")
cfo.close()
lock.release()
#print cluster + ": " + struct
#stdout.flush()
except Exception, e:
cluster, struct, "\nERROR!"
stdout.flush()
lock.release()
def create_final_output(groupfasta, basefolder, minseqs=1, cpus=1):
'''Function for multithreading. Creates the final BayesFold alignment and
writes to files, then r2r struct and infernal CM file'''
# skip if already run and program just crashed or whatever
currgroup = groupfasta.split("/")[-1].split(".")[0]
currotufolder = basefolder + currgroup
if exists(currotufolder):
return
# load seqs and make sure we have enough
aln = LoadSeqs(groupfasta, moltype=RNA, aligned=True)
count = count_seqs(aln.Names)
if count < minseqs:
return
# get weights for each sequence. weight==count
weights = []
maxweight = 0
for header in aln.Names:
weight = count_seqs(header)
if weight > maxweight:
maxweight = weight
weights.append(header.split()[0])
weights.append(str(weight))
# fold alignment with bayesfold
aln, struct = bayesfold(aln, align=False)
# write log information
mkdir(currotufolder)
with open(currotufolder + "/log.txt", 'w') as logout:
logout.write(' '.join([
currgroup, ":\n",
str(count), "sequences\n",
str(aln.getNumSeqs()), "unique sequences\nStructure: ", struct,
"\n"
]))
# write out alignment and structure in fasta format
with open(currotufolder + "/bayesfold-aln.fasta", 'w') as alnout:
alnout.write(">SS_cons\n%s\n%s" % (struct, aln.toFasta()))
# shave off info in header for stockholm
aln = LoadSeqs(data=aln, moltype=RNA, label_to_name=lambda x: x.split()[0])
# create stockholm formatted alignment
sto = stockholm_from_alignment(aln, GC_annotation={'SS_cons': struct})
del aln
# create standard weights for infernal
infweights = ""
for pos in range(0, len(weights), 2):
infweights = ''.join([
infweights,
'# =GS %s WT %s\n' %
(weights[pos], str(float(weights[pos + 1]) / maxweight))
])
# create weights for r2r
r2r_weights = "# =GF USE_THIS_WEIGHT_MAP " + ' '.join(weights)
# create sto file with r2r and std weights
sto = sto.split("\n")
sto[-1] = infweights.strip()
sto.append(r2r_weights)
sto.append("//\n")
stofile = currotufolder + "/bayesfold-aln.sto"
with open(stofile, 'w') as alnout:
alnout.write('\n'.join(sto))
# make R2R secondary structure for alignment
make_r2r(stofile, currotufolder, currgroup)
# create CM file for infernal from group
cmbuild_from_file(stofile,
currotufolder + "/cmfile.cm",
params={'--wgiven': True})
calibrate_cmfile(currotufolder + "/cmfile.cm", cpus=cpus)
