def write_ensemble_metrics(row, id_field):
bp_filename = 'bpps_%s.npy.gz' % str(row[id_field])
# bpps exist already
if os.path.exists(bp_filename):
f = gzip.GzipFile(bp_filename, "r")
bpp_mat = np.load(f)
print('loaded ', row[id_field])
else: # calculate bpps and cache
bpp_mat = bpps(row['sequence'], package='eternafold')
print("wrote ", row[id_field])
# cache
f = gzip.GzipFile(bp_filename, 'w')
np.save(file=f, arr=bpp_mat)
f.close()
punp_vector = 1-np.sum(bpp_mat, axis=0)
aup = np.mean(punp_vector)
sup_init = np.sum(punp_vector[:14])
return aup, sup_init
def threshknot_util(sequence, package='vienna_2', theta=0):
'''
Inputs:
sequence: RNA sequence
package: folding package to use
Set theta = 0 to not filter base pairs as in ThreshKnot.
Returns: N x N matrix of base pair probabilities. Nonzero entries represent base pairs
predicted in final (possibly pseudoknotted) structure.
Probabilities are their associated probability (obvs).
'''
bp_matrix = bpps(sequence, package=package)
# if desired, filter base pair probabilities below a cutoff
bp_matrix[np.where(bp_matrix <= theta)] = 0
output = np.zeros([len(sequence), len(sequence)])
# ProbKnot heuristic part 1: get all base pairs where p(ij) == p_max(i)
output[np.where(bp_matrix == np.max(bp_matrix, axis=0))] = 1
# ProbKnot heuristic part 2: get all base pairs where p(ij) == p_max(j)
array_of_bps = np.clip(output + np.transpose(output) - 1, 0, 1)
# setting all bp probabilities not corresponding to a final selected base pair to zero
bp_matrix[np.where(array_of_bps == 0)] = 0
return bp_matrix
def get_secstruct_mea(int_start, int_end, ref_seq, secstruct_interval=20):
sequence = ref_seq[(int_start-secstruct_interval):(int_end+secstruct_interval)]
bp_matrix = bpps(sequence, package='contrafold_2')
best_struct = ''
best_mcc = 0
for log_gamma in range(-2,2):
mea_mdl = MEA(bp_matrix,gamma=10**log_gamma)
[exp_sen, exp_ppv, exp_mcc, exp_fscore] = mea_mdl.score_expected()
if exp_mcc > best_mcc:
best_struct = mea_mdl.structure
best_mcc = exp_mcc
conserved_str = '.'*secstruct_interval + '*'*(int_end - int_start) + '.'*secstruct_interval
return((best_struct, best_mcc, conserved_str, sequence))
from arnie.utils import write_vector_to_file
if __name__=='__main__':
p = argparse.ArgumentParser(description=
"""Write unpaired posterior probabilities to files.
""")
p.add_argument("seq_dir", nargs='+',
help="path to dir of *.seq files")
p.add_argument("-o", help="name of output dir")
p.add_argument("-p", "--package", default='vienna_2', help="Package to use")
if len(sys.argv)==1:
p.print_help(sys.stderr)
sys.exit(1)
args = p.parse_args()
if not os.path.exists('./%s' % args.o):
os.makedirs('./%s' % args.o)
for seqfile in args.seq_dir:
print(seqfile)
seq=open(seqfile,'r').readlines()[-1].rstrip()
seq_id = os.path.basename(seqfile).replace('.seq','')
unp_vector = 1-np.sum(bpps.bpps(seq, package=args.package),axis=0)
with open("%s/%s.unp" % (args.o, seq_id),'w') as f:
write_vector_to_file(unp_vector, f)
from arnie.utils import write_matrix_to_file
if __name__=='__main__':
p = argparse.ArgumentParser(description=
"""
Write base pairing probability matrices to files.
""")
p.add_argument("seq_dir", nargs='+',
help="path to dir of *.seq files")
p.add_argument("-o", help="name of output dir")
p.add_argument("-p", "--package", default='vienna_2',
help="Package to use")
if len(sys.argv)==1:
p.print_help(sys.stderr)
sys.exit(1)
args = p.parse_args()
if not os.path.exists('./%s' % args.o):
os.makedirs('./%s' % args.o)
for seqfile in args.seq_dir:
print(seqfile)
seq=open(seqfile,'r').readlines()[-1].rstrip()
seq_id = os.path.basename(seqfile).replace('.seq','')
bp_matrix = bpps.bpps(seq, package=args.package)
with open("%s/%s.bpps" % (args.o, seq_id),'w') as f:
write_matrix_to_file(bp_matrix, f)
def test_bpps(pkg):
p = bpps(sample_seq, package=pkg)
print('test bpps %s' % pkg)
print(p[0])
return
def calc_bpp(seq):
return bpps(seq, package=args.package, T=args.temp)