def _ui_callback(self, descr, pdb, cid, seq_marker):
if descr not in self.matchers:
error_msg = f"{descr} not found in list of matches. " \
f"This is an internal error."
raise Exception(error_msg)
return_val = calculate_single(pdb, cid, seq_marker)
p_all_sno = self.matchers[descr].query(return_val)
return_val_sno = return_val.sno.values
residues = return_val.res
res_to_AA1 = []
for res in residues.values:
if res in AA3_to_AA1:
res_to_AA1.append(AA3_to_AA1[res])
else:
res_to_AA1.append('X')
output = dict()
output['residues'] = res_to_AA1
alphabets = []
for res in residues:
if res in AA3_to_AA1:
alphabets.append(res)
else:
alphabets.append("TYR")
output['alphabet'] = alphabets
output['per_point'] = dict()
for term, values in p_all_sno.items():
output['per_point'][return_val_sno[term]] = values
output['generic'] = defaultdict(dict)
for descr_local, matcher in self.matchers_generic.items():
p_all_sno = matcher.query(return_val)
return_val_sno = return_val.sno.values
for term, values in p_all_sno.items():
output['generic'][descr_local][return_val_sno[term]] = values
try:
segmented_scores = self.segment_calculater.query(
descr, pdb, cid, seq_marker)
except:
print(f"segment_calculater.query failed with error"
f" {traceback.format_exc()}")
output['segment'] = None
else:
output['segment'] = segmented_scores
return output
def _ui_callback(pdb, cid, seq_marker, matcher):
return_val = calculate_single(pdb, cid, seq_marker)
p_all_sno = matcher.query(return_val)
return_val_sno = return_val.sno.values
output = dict()
output['per_point'] = dict()
for term, values in p_all_sno.items():
output['per_point'][return_val_sno[term]] = values
return 0, output['per_point']
# pdb = '1a72'
# cid = 'A'
# seq_marker = '187'
# matcher = get_matcher()
#
# ret_code, output = _ui_callback(pdb, cid, seq_marker, matcher)
#
# score_map = [dict() for __ in range(len(output))]
# top_candidates = set()
# top_5_scores = np.ones(len(output), dtype=float)
#
#
# for i, value in enumerate(output.values()):
# for j, (identifier, score) in enumerate(value):
# id_str = (identifier[0], identifier[1], identifier[2])
# if j < 10:
# top_candidates.add(id_str)
# if j < 30:
# top_5_scores[i] = min(top_5_scores[i], score)
# score_map[i][id_str] = score
#
# segment_length = 5
# best_score_arr = []
#
# for i in range(len(output) - segment_length):
# max_score, max_cand = 0, None
# for candidate in top_candidates:
# cand_score = 0
# for j in range(segment_length):
# try:
# cand_score += score_map[i+j][candidate]
# except KeyError:
# continue
# if cand_score > max_score:
# max_score = cand_score
# max_cand = candidate
# best_score_arr.append([max_score, max_cand])
#
# best_i = np.argmax([i[0] for i in best_score_arr])
# best_candidate = best_score_arr[best_i][1]
# # extend on both ends
# # to the left
# left_i, right_i = best_i, best_i+segment_length
# while left_i >= 0:
# if score_map[left_i][best_candidate] < top_5_scores[left_i]:
# break
# left_i -= 1
#
# while right_i < len(output):
# if score_map[right_i][best_candidate] < top_5_scores[right_i]:
# break
# right_i += 1
#
# candidate_output = [None for __ in range(len(output))]
# for i in range(left_i, right_i+1):
# candidate_output[i] = [best_candidate, score_map[i][best_candidate],
# best_score_arr[best_i][0]]
#
# segment_length = 3
#
# while left_i >= segment_length:
# seg_left, seg_right = left_i - segment_length, left_i
# max_score, max_cand = 0, None
# for candidate in top_candidates:
# cand_score = 0
# for j in range(seg_left, seg_right):
# try:
# cand_score += score_map[j][candidate]
# except KeyError:
# continue
# if cand_score > max_score:
# max_score = cand_score
# max_cand = candidate
#
# while seg_left >= 0:
# if score_map[seg_left][max_cand] < top_5_scores[seg_left]:
# break
# seg_left -= 1
# seg_left = max(seg_left, 0)
# for i in range(seg_left, seg_right):
# candidate_output[i] = [max_cand, score_map[i][max_cand], max_score]
# left_i = seg_left
#
# while right_i <= len(output) - segment_length:
# seg_left, seg_right = right_i, right_i+segment_length
# max_score, max_cand = 0, None
# for candidate in top_candidates:
# cand_score = 0
# for j in range(seg_left, seg_right):
# try:
# cand_score += score_map[j][candidate]
# except KeyError:
# continue
# if cand_score > max_score:
# max_score = cand_score
# max_cand = candidate
#
# while seg_right < len(output):
# if score_map[seg_right][max_cand] < top_5_scores[seg_right]:
# break
# seg_right += 1
#
# for i in range(seg_left, seg_right):
# candidate_output[i] = [max_cand, score_map[i][max_cand], max_score]
# right_i = seg_right
#
# for term in candidate_output:
# print(term)
def query(self, descr, pdb, cid, seq_marker):
matcher = self.matchers[descr]
try:
return_val = calculate_single(pdb, cid, seq_marker)
except:
msg = f"query failed with error {traceback.format_exc()}"
raise Exception(msg)
p_all_sno = matcher.query(return_val)
sno_list = return_val.sno.values
sno_probs = dict()
for relative_sno, probs in p_all_sno.items():
actual_sno = sno_list[relative_sno]
sno_probs[actual_sno] = probs
descr_length = len(sno_probs)
score_map = [dict() for __ in range(descr_length)]
top_candidates = set()
top_n_scores = np.ones(descr_length, dtype=float)
for i, probs in enumerate(sno_probs.values()):
for j, (identifier, score) in enumerate(probs):
id_str = (identifier[0], identifier[1], identifier[2])
if j < self.num_top_candidates:
top_candidates.add(id_str)
if j < self.extension_cutoff:
top_n_scores[i] = min(top_n_scores[i], score)
score_map[i][id_str] = score
best_scores = []
for i in range(descr_length - self.segment_len):
max_score, max_cand = 0, None
for candidate in top_candidates:
cand_score = 0
for j in range(self.segment_len):
try:
cand_score += score_map[i + j][candidate]
except KeyError:
# skip these candidates
continue
if cand_score > max_score:
max_score = cand_score
max_cand = candidate
best_scores.append([max_score, max_cand])
best_i = np.argmax([i[0] for i in best_scores])
# to dismiss ide inspection error
best_i = int(best_i)
best_candidate = best_scores[best_i][1]
# extend on both ends
# to the left
left_i, right_i = best_i, best_i + self.segment_len
while left_i > 0:
if score_map[left_i][best_candidate] < top_n_scores[left_i]:
break
left_i -= 1
# to the right
while right_i < descr_length - 1:
if score_map[right_i][best_candidate] < top_n_scores[right_i]:
break
right_i += 1
candidate_output = [[] for __ in range(descr_length)]
for i in range(left_i, right_i + 1):
candidate_output[i] = [
best_candidate, score_map[i][best_candidate],
best_scores[best_i][0]
]
while left_i >= self.extension:
seg_left, seg_right = left_i - self.extension, left_i
max_score, max_cand = 0, None
for candidate in top_candidates:
cand_score = 0
for j in range(seg_left, seg_right):
try:
cand_score += score_map[j][candidate]
except KeyError:
continue
if cand_score > max_score:
max_score = cand_score
max_cand = candidate
while seg_left >= 0:
if seg_left in score_map and max_cand in score_map[seg_left] \
and \
score_map[seg_left][max_cand] < top_n_scores[seg_left]:
break
seg_left -= 1
seg_left = max(seg_left, 0)
for i in range(seg_left, seg_right):
if max_cand in score_map[i]:
score_to_add = score_map[i][max_cand]
else:
score_to_add = 0.01
candidate_output[i] = [max_cand, score_to_add, max_score]
left_i = seg_left
while right_i <= descr_length - self.extension:
seg_left, seg_right = right_i, right_i + self.extension
max_score, max_cand = 0, None
for candidate in top_candidates:
cand_score = 0
for j in range(seg_left, seg_right):
try:
cand_score += score_map[j][candidate]
except KeyError:
continue
if cand_score > max_score:
max_score = cand_score
max_cand = candidate
while seg_right < descr_length:
if score_map[seg_right][max_cand] < top_n_scores[seg_right]:
break
seg_right += 1
for i in range(seg_left, seg_right):
candidate_output[i] = [
max_cand, score_map[i][max_cand], max_score
]
right_i = seg_right
mapped_output = dict()
for sno, value in zip(sno_probs.keys(), candidate_output):
mapped_output[sno] = value
return mapped_output