def get_design_data():
"""
:return: goes over all designed coh-doc pairs and returns a DF similar to the one used to create the decision tree
"""
dsn_cohs = read_multi_fastas(design_data_root+'all_designed_cohs.fasta', suffix_to_remove='.')
dsn_docs = read_multi_fastas(design_data_root+'all_designed_docs.fasta', suffix_to_remove='.')
df_ = pd.DataFrame(columns=columns, index=range(1, len(list(dsn_cohs.keys()))))
interface_positions = parse_interface_positions()
for i, doc_seq in enumerate(list(dsn_docs.values())):
coh_seq = dsn_cohs[doc_seq.name]
doc_model = doc_seq.name.split('A_')[1].split('_')[0]
coh_identities = {typ: coh_seq[pos] for typ, pos in interface_positions['coh']['1ohz'].items()}
doc_identities = {typ: doc_seq[pos] for typ, pos in interface_positions['doc'][doc_model].items()}
coh_core = [core_res_to_identity(coh_identities[v], 'coh') for v in ordered_positions['coh'] if 'core' in v]
coh_rim = [rim_res_to_type_binary(coh_identities[v]) for v in ordered_positions['coh'] if 'rim' in v]
doc_core = [core_res_to_identity(doc_identities[v], 'doc') for v in ordered_positions['doc'] if 'core' in v]
doc_rim = [rim_res_to_type_binary(doc_identities[v]) for v in ordered_positions['doc'] if 'rim' in v]
coh_core_list, coh_rim_list = [], []
[coh_core_list.append(a) for b in coh_core for a in b]
[coh_rim_list.append(a) for b in coh_rim for a in b]
doc_core_list, doc_rim_list = [], []
[doc_core_list.append(a) for b in doc_core for a in b]
[doc_rim_list.append(a) for b in doc_rim for a in b]
df_.loc[i+1] = [coh_seq.name, doc_seq.name, 0, 0] + coh_core_list + coh_rim_list + \
doc_core_list + doc_rim_list + [None]
return df_
def parse_binding_data() -> pd.DataFrame:
"""
:return: data frame 'coh_name', 'doc_name', 'coh_seq', 'doc_seq', 'binders' for Rachel's data
"""
from _binding_data import binding_data
rachel_root = '/home/labs/fleishman/jonathaw/decision_tree/'
cohs = read_multi_fastas(rachel_root+'cohesins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/', lower=True)
docs = read_multi_fastas(rachel_root+'dockerins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/', lower=True)
rachel_bind = binding_data()
vered_bind = parse_vered_binding()
result = pd.DataFrame(columns=['coh_name', 'doc_name', 'coh_seq', 'doc_seq', 'binders'])
i = 1
for coh, docs_dict in rachel_bind.items():
for doc, res in docs_dict.items():
result.loc[i] = [coh, doc, cohs[coh], docs[doc], rachel_bind[coh][doc]]
i += 1
for coh, docs_dict in vered_bind.items():
for doc, res in docs_dict.items():
result.loc[i] = [coh, doc, cohs[coh], docs[doc], vered_bind[coh][doc] == 1]
i += 1
for name in ['1ohz', '2b59', '2ozn', '2vn5', '2y3n', '3ul4', '4fl4', '4fl5', '4dh2', '4uyp', '5new']:
result.loc[i] = [name, name, cohs[name], docs[name], True]
i += 1
print('there are %i rows in the data' % (i-1))
return result
def analyse_cliques(cliques):
"""
:param cliques: list of cliques
:return: prints an anlysis
"""
coh_seqs = read_multi_fastas('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/reclique_18Nov/stabilisation/all_stabilised/all_j_st_cohs.fasta', '_st.A')
doc_seqs = read_multi_fastas('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/reclique_18Nov/stabilisation/all_stabilised/all_j_st_docs.fasta', '_st.B')
cliques_by_charges = parse_cliques_lists('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/reclique_18Nov/stabilisation/all_stabilised/cliques_2_1.txt', remove='_st')
original_names = parse_name_translation('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/clique_6_pdbs/mini_diagonal_11Nov/minidiagonal_pdbs/translate_names.txt')
clqs_by_len = {k: [] for k in range(1, 8)}
for clq in cliques:
clqs_by_len[len(clq)].append(clq)
designs_in_all_clqs = []
for length in range(10, 6, -1):
if length not in clqs_by_len.keys():
continue
for clq in clqs_by_len[length]:
coh_diffs, doc_diffs, doc_diffs_symm, doc_bb_diffs = [], [], [], []
print('\n\n\nfor clq', clq)
for mem1 in clq:
designs_in_all_clqs.append(mem1)
wt_doc = original_names[mem1[1]+'.pdb.gz'][10:14]
res1 = Result(mem1, coh_seqs[mem1[0]], doc_seqs[mem1[1]], 1, j=True, doc_wt=wt_doc)
res1_doc_symm = switch_symm_changer(res1.doc_switch)
for mem2 in clq:
if mem1 != mem2:
wt_doc = original_names[mem2[1]+'.pdb.gz'][10:14]
res2 = Result(mem2, coh_seqs[mem2[0]], doc_seqs[mem2[1]], 1, j=True, doc_wt=wt_doc)
doc_bb_diffs.append(are_docs_from_diff_clusters(res1.doc_wt, res2.doc_wt))
coh_diffs.append(switches_differ({}, res1.coh_switch, res2.coh_switch))
doc_diffs.append(switches_differ({}, res1.doc_switch, res2.doc_switch))
doc_diffs_symm.append(switches_differ({}, res1_doc_symm, res2.doc_switch))
print('results', res1)
print('results', res2)
print('docs diff', switches_differ({}, res1.doc_switch, res2.doc_switch))
print('doc symm diff', switches_differ({}, res1_doc_symm, res2.doc_switch))
print('doc BB dif', are_docs_from_diff_clusters(res1.doc_wt, res2.doc_wt))
print('coh diff', switches_differ({}, res1.coh_switch, res2.coh_switch))
N
print('for clq %r found the following results:' % clq)
print('doc_bb_diffs', doc_bb_diffs)
print('doc_diffs', doc_diffs)
print('doc_diffs_symm', doc_diffs_symm)
print('coh_diffs', coh_diffs)
print('total', sum([1 for a in doc_bb_diffs if a] + doc_diffs + doc_diffs_symm + coh_diffs))
all_cohs = list(set([a[0] for a in designs_in_all_clqs]))
all_docs = list(set([a[1] for a in designs_in_all_clqs]))
print('these are all the cohs i need: %s, total %i' % (', '.join(all_cohs), len(all_cohs)))
print('these are all the docs i need: %s, total %i' % (', '.join(all_docs), len(all_docs)))
print('LONGEST CLIQUES FOUND ARE %i' % max([len(clq) for clq in cliques]))
coh_doc_purples = creat_coh_doc_purples()
for clq in clqs_by_len[max(list(clqs_by_len.keys()))]:
print('clq', clq)
cohs = [a[0] for a in clq]
docs = [a[1] for a in clq]
df = pd.DataFrame(index=docs, columns=cohs)
for coh in cohs:
for doc in docs:
df[coh][doc] = coh_doc_purples[coh][doc]
show_prediction_heat_map(df)
def retrive_relevant_poses() -> (dict, dict):
"""
:return: seq dicts for cohs and docs, holding only the relevqant positions, determined by 1OHZ
"""
cohs_old = read_multi_fastas(root_path + 'cohesins_from_rachel.fasta_aln',
suffix_to_remove='/')
docs_old = read_multi_fastas(root_path + 'dockerins_from_rachel.fasta_aln',
suffix_to_remove='/')
coh_1ohz = cohs_old['1OHZ']
coh_poses = [
coh_1ohz.non_aligned_position_at_aligned(p) for p in coh_poses_1ohz
]
doc_1ohz = docs_old['1OHZ']
doc_poses = [
doc_1ohz.non_aligned_position_at_aligned(p) for p in doc_poses_1ohz
]
cohs_new, docs_new = {}, {}
for coh, res in cohs_old.items():
cohs_new[coh] = AASeq(string=''.join(
res.get_aligned_positions(coh_poses)),
name=coh)
for doc, res in docs_old.items():
docs_new[doc] = AASeq(string=''.join(
res.get_aligned_positions(doc_poses)),
name=doc)
return cohs_new, docs_new
def post_pred_cliques(args):
run_filters = generate_run_filters(args={'ddg': 25.0, 'sasa': 1400, 'shape': 0.6, 'packstat': 0.6, 'buried_2': 3})
if not os.path.isfile('./all_data.obj'):
sc_files = [a for a in os.listdir('./') if '.score' in a]
cohs_seqs = read_multi_fastas('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/all_cohs.fasta')
docs_seqs = read_multi_fastas('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/all_docs.fasta')
results = []
for sc_file in sc_files:
seq_name = '_'.join(sc_file.split('_')[1:8])
coh_name = seq_name+'.pdb.gz.A'
doc_name = seq_name+'.pdb.gz.B'
sc_dict = score2dict(sc_file)
ynum = re.search('y[0-9]{3}', sc_file).group(0)
passed, failed = all_who_pass_run_filters(args, sc_dict, run_filters)
if len(passed) >= args['purples_threshold']:
r = Result(seq_name, cohs_seqs[coh_name], docs_seqs[doc_name], len(passed))
results.append(r)
with open('./all_data.obj', 'wb') as fout:
pickle.dump(results, fout)
else:
with open('./all_data.obj', 'rb') as fin:
results = pickle.load(fin)
if not os.path.isfile('./graph.obj'):
result_dict = {i+1: r for i, r in enumerate(results)}
G = nx.Graph()
[G.add_node(a) for a in result_dict.keys()]
for n1 in G.nodes_iter():
for n2 in G.nodes_iter():
if n1 != n2:
coh_sw_1, coh_sw_2 = result_dict[n1].coh_switch, result_dict[n2].coh_switch
doc_sw_1, doc_sw_2 = result_dict[n1].doc_switch, result_dict[n2].doc_switch
doc_wt_1, doc_wt_2 = result_dict[n1].doc_wt, result_dict[n2].doc_wt
doc_diff = 1 if are_docs_from_diff_clusters(doc_wt_1, doc_wt_2) else 0
symm_switch = switch_symm_changer(doc_sw_2)
if switches_differ({'diff_by': args['diff_by']}, coh_sw_1, coh_sw_2) >= args['diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, doc_sw_2) + doc_diff >= args['doc_diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, symm_switch) + doc_diff >= args['doc_diff_by']:
G.add_edge(n1, n2)
print('adding edge\n', result_dict[n1], '\n', result_dict[n2])
else:
print('NOT\n', result_dict[n1], '\n', result_dict[n2])
cliques = [a for a in nx.find_cliques(G)]
max_len = max([len(a) for a in cliques])
max_cliques = [a for a in cliques if len(a) == max_len]
for clq in max_cliques:
print(clq, '\n', '\n'.join([str(result_dict[a]) for a in clq]))
def main():
pd.set_option('display.max_columns', 500)
pd.set_option('display.width', 1000)
parser = argparse.ArgumentParser()
parser.add_argument('-gremlin_file',
type=str,
help='full path to Gremlin output')
parser.add_argument('-MSA', type=str, help='full path the MSA file')
parser.add_argument('-probability_threshold',
type=float,
help='threshold above which probability is considered')
parser.add_argument('-query_name',
type=str,
help='query name as it is written in the MSA')
args = vars(parser.parse_args())
gremlin = parse_gremlin(args['gremlin_file'],
args['probability_threshold'])
msa = read_multi_fastas(args['MSA'], add_aligned=True)
with open(args['query_name'] + '.gssm', 'w+') as fout:
for k, v in gremlin.items():
iden_frq = create_identitiy_frequency_df(k, msa,
args['query_name'])
fout.write('pos_1 %i pos_2 %i probability %f\n' % (k[0], k[1], v))
fout.write(str(iden_frq) + '\n')
def validate(args):
# original_seqs = read_multi_fastas(args['original_seqs_file'], suffix_to_remove='_')
DNA_seqs = read_multi_fastas(args['DNA_seqs_file'], suffix_to_remove='.')
for k, v in DNA_seqs.items():
# assert original_seqs[k].get_seq() in DNA2AA(v.get_seq())
if not gen9_standards(v.get_seq):
print('seq name %s does not comply with Gen9 standards' % k)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-coh_files', nargs='+')
parser.add_argument('-doc_files', nargs='+')
parser.add_argument('-threshold', type=float, default=30.0)
args = vars(parser.parse_args())
cohs = OrderedDict({
k: v
for a in args['coh_files'] for k, v in read_multi_fastas(a).items()
})
docs = OrderedDict({
k: v
for a in args['doc_files'] for k, v in read_multi_fastas(a).items()
})
mw_df = pd.DataFrame(columns=['coh_seq', 'coh_weight'] + list(docs.keys()))
for i, coh in enumerate(cohs.values()):
coh.add_prefix = flanks['coh']['cbm']
coh_weight = coh.calc_molecular_weight()
print(coh.get_seq(), coh_weight)
weights_combined = []
for doc in docs.values():
doc.add_prefix(flanks['doc']['xyn'])
weights_combined.append(coh_weight + doc.calc_molecular_weight())
mw_df.loc[coh.name] = [coh, coh_weight] + weights_combined
print(mw_df)
diffs = []
for coh1 in cohs.keys():
for doc1 in docs.keys():
coh_doc_1 = mw_df[doc1][coh1]
for coh2 in cohs.keys():
for doc2 in docs.keys():
if coh1 != coh2 and doc1 != doc2:
diff = abs(coh_doc_1 - mw_df[doc2][coh2])
diffs.append(diff)
if diff <= args['threshold']:
print(
'%s %s and %s %s have a weight difference of only %f'
% (coh1, doc1, coh2, doc2, diff))
plt.boxplot(diffs)
plt.show()
def retrive_relevant_poses() -> (dict, dict):
"""
:return: seq dicts for cohs and docs, holding only the relevqant positions, determined by 1OHZ
"""
cohs_old = read_multi_fastas(root_path+'cohesins_from_rachel.fasta_aln', suffix_to_remove='/')
docs_old = read_multi_fastas(root_path+'dockerins_from_rachel.fasta_aln', suffix_to_remove='/')
coh_1ohz = cohs_old['1OHZ']
coh_poses = [coh_1ohz.non_aligned_position_at_aligned(p) for p in coh_poses_1ohz]
doc_1ohz = docs_old['1OHZ']
doc_poses = [doc_1ohz.non_aligned_position_at_aligned(p) for p in doc_poses_1ohz]
cohs_new, docs_new = {}, {}
for coh, res in cohs_old.items():
cohs_new[coh] = AASeq(string=''.join(res.get_aligned_positions(coh_poses)), name=coh)
for doc, res in docs_old.items():
docs_new[doc] = AASeq(string=''.join(res.get_aligned_positions(doc_poses)), name=doc)
return cohs_new, docs_new
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-coh_files', nargs='+')
parser.add_argument('-doc_files', nargs='+')
parser.add_argument('-threshold', type=float, default=30.0)
args = vars(parser.parse_args())
cohs = OrderedDict({k: v for a in args['coh_files'] for k, v in read_multi_fastas(a).items()})
docs = OrderedDict({k: v for a in args['doc_files'] for k, v in read_multi_fastas(a).items()})
mw_df = pd.DataFrame(columns=['coh_seq', 'coh_weight']+list(docs.keys()))
for i, coh in enumerate(cohs.values()):
coh.add_prefix = flanks['coh']['cbm']
coh_weight = coh.calc_molecular_weight()
print(coh.get_seq(), coh_weight)
weights_combined = []
for doc in docs.values():
doc.add_prefix(flanks['doc']['xyn'])
weights_combined.append(coh_weight+doc.calc_molecular_weight())
mw_df.loc[coh.name] = [coh, coh_weight] + weights_combined
print(mw_df)
diffs = []
for coh1 in cohs.keys():
for doc1 in docs.keys():
coh_doc_1 = mw_df[doc1][coh1]
for coh2 in cohs.keys():
for doc2 in docs.keys():
if coh1 != coh2 and doc1 != doc2:
diff = abs(coh_doc_1 - mw_df[doc2][coh2])
diffs.append(diff)
if diff <= args['threshold']:
print('%s %s and %s %s have a weight difference of only %f' % (coh1, doc1, coh2, doc2, diff))
plt.boxplot(diffs)
plt.show()
def main():
pd.set_option('display.max_columns', 500)
pd.set_option('display.width', 1000)
parser = argparse.ArgumentParser()
parser.add_argument('-gremlin_file', type=str, help='full path to Gremlin output')
parser.add_argument('-MSA', type=str, help='full path the MSA file')
parser.add_argument('-probability_threshold', type=float, help='threshold above which probability is considered')
parser.add_argument('-query_name', type=str, help='query name as it is written in the MSA')
args = vars(parser.parse_args())
gremlin = parse_gremlin(args['gremlin_file'], args['probability_threshold'])
msa = read_multi_fastas(args['MSA'], add_aligned=True)
with open(args['query_name']+'.gssm', 'w+') as fout:
for k, v in gremlin.items():
iden_frq = create_identitiy_frequency_df(k, msa, args['query_name'])
fout.write('pos_1 %i pos_2 %i probability %f\n' % (k[0], k[1], v))
fout.write(str(iden_frq) + '\n')
parser.add_argument('-coh_seqs_file')
parser.add_argument('-doc_seqs_file')
parser.add_argument('-mode')
parser.add_argument('-n', type=int, default=1)
parser.add_argument('-diff_by', type=int, default=2)
parser.add_argument('-doc_diff_by', type=int, default=1)
parser.add_argument('-score_dir', type=str, default='./')
parser.add_argument('-purples_threshold', type=int, default=50)
args = vars(parser.parse_args())
if args['mode'] != 'bins_diagonal' and args[
'mode'] != 'post_pred_cliques' and args[
'mode'] != 'minidiagonal_cliques':
scores = score2dict(args['score_file'])
run_filters = '' #generate_run_filters()
coh_seq_dict = read_multi_fastas(args['coh_seqs_file'],
suffix_to_remove='.pdb.gz')
doc_seq_dict = read_multi_fastas(args['doc_seqs_file'],
suffix_to_remove='.pdb.gz')
if args['mode'] == 'switches_n_cliques':
switches, num_bins = make_switches(args, scores, run_filters,
coh_seq_dict)
max_cliques = best_cliques(args, list(switches.keys()))
with open('switches.obj', 'wb') as sw_file:
pickle.dump(switches, sw_file)
with open('max_cliques.obj', 'wb') as clq_file:
pickle.dump(max_cliques, clq_file)
elif args['mode'] == 'choose_by_identity':
with open('switches.obj', 'rb') as sw_in:
def add_primers_to_all(args):
DNA_seqs = read_multi_fastas(args['DNA_seqs_file'])
for k, v in DNA_seqs.items():
print('>%s' % k)
print(add_primers(v.get_seq, args['type']))
def validate_data_frame(data_df: pd.DataFrame, prepared_df: pd.DataFrame) -> None:
"""
:param data_df: binding data frame
:param prepared_df: binary data frame
:return: prints if there is something wrong...
"""
rachel_root = '/home/labs/fleishman/jonathaw/decision_tree/'
cohs = read_multi_fastas(rachel_root+'cohesins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/')
docs = read_multi_fastas(rachel_root+'dockerins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/')
# coh_1ohz = cohs['1OHZ']
# doc_1ohz = docs['1OHZ']
coh_crys_seqs = [c for c in cohs.values() if c.name in ['1ohz', '2b59', '2ozn', '2vn5', '2y3n', '3ul4',
'4fl4', '4fl5', '4dh2', '4uyp', '5new']]
doc_crys_seqs = [d for d in docs.values() if d.name in ['1ohz', '2b59', '2ozn', '2vn5', '2y3n', '3ul4',
'4fl4', '4fl5', '4dh2', '4uyp', '5new']]
# coh_poses = [coh_1ohz.non_aligned_position_at_aligned(p) for p in coh_poses_1ohz]
# doc_poses = [doc_1ohz.non_aligned_position_at_aligned(p) for p in doc_poses_1ohz]
features = list(prepared_df.columns[4:-1])
interface_positions = parse_interface_positions()
coh_poses = {coh: {typ: cohs[coh].non_aligned_position_at_aligned(pos) for typ, pos in typos.items()} for coh, typos
in interface_positions['coh'].items()}
doc_poses = {doc: {typ: docs[doc].non_aligned_position_at_aligned(pos) for typ, pos in typos.items()} for doc, typos
in interface_positions['doc'].items()}
for i in range(1, len(data_df.index)):
# i = len(data_df.index)
print('i is %i' % i)
print(data_df.loc[i])
if data_df.loc[i]['coh_name'] != prepared_df.loc[i]['coh_name'] or \
data_df.loc[i]['doc_name'] != prepared_df.loc[i]['doc_name']:
print('not the same names', data_df.loc[i]['doc_name'], prepared_df.loc[i]['doc_name'])
sys.exit()
coh_seq = data_df.loc[i]['coh_seq']
# coh_q_poses = coh_seq.get_aligned_positions(coh_poses)
doc_seq = data_df.loc[i]['doc_seq']
# doc_q_poses = doc_seq.get_aligned_positions(doc_poses)
prepared_row = row_to_dict(prepared_df.loc[i])
similar_coh, coh_iden = highest_seq_similarity(coh_crys_seqs, data_df.loc[i]['coh_seq'])
similar_doc, doc_iden = highest_seq_similarity(doc_crys_seqs, data_df.loc[i]['doc_seq'])
coh_identities = {typ: data_df.loc[i]['coh_seq'].get_aligned_positions([pos])[0] for typ, pos in
coh_poses[similar_coh.name].items()}
doc_identities = {typ: data_df.loc[i]['doc_seq'].get_aligned_positions([pos])[0] for typ, pos in
doc_poses[similar_doc.name].items()}
# for pos in positions['core_coh']:
# if coh_q_poses[pos] != prepared_row['coh_core_%i' % pos]:
# print('not the same coh query pos differs from row', pos, coh_q_poses[pos], prepared_row['coh_core_%i' % pos])
# sys.exit()
# for pos in positions['core_doc']:
# if doc_q_poses[pos] != prepared_row['doc_core_%i' % pos]:
# print('not the same doc query pos differs from row')
# sys.exit()
# for pos in positions['rim_coh']:
# if [k for k, v in type_to_res.items() if coh_q_poses[pos] in v][0] != prepared_row['coh_rim_%i' % pos] and \
# not ([k for k, v in type_to_res.items() if coh_q_poses[pos] in v][0] == 'NA' and
# prepared_row['coh_rim_%i' % pos] == '-'):
# print('breaking', [k for k, v in type_to_res.items() if coh_q_poses[pos] in v][0],
# prepared_row['coh_rim_%i' % pos])
# sys.exit()
for fea in features:
if prepared_df.loc[i][fea] not in [0, 1]:
print('found problem at row', i, prepared_df.loc[i][fea])
# break
print('your df is validated')
def minidiagonal_cliques(args):
original_names = parse_name_translation('/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/clique_6_pdbs/mini_diagonal_11Nov/minidiagonal_pdbs/translate_names.txt')
coh_seqs = read_multi_fastas(args['coh_seqs_file'], suffix_to_remove='.A')
doc_seqs = read_multi_fastas(args['doc_seqs_file'], suffix_to_remove='.B')
if not os.path.isfile('all_results.obj') or not os.path.isfile('all_bins.obj'):
print('creating bins and results')
all_results, bins = {}, {}
for design in coh_seqs.keys():
r = Result(design, coh_seqs[design], doc_seqs[design], 0, j=True, originals=original_names)
if not 2 < r.coh_switch.count('n') + r.coh_switch.count('p') < 8 and not 2 <= r.coh_switch.count('p') <= 3:
continue
all_results[design] = r
d_sw = r.coh_switch+'-'+r.doc_switch+'-'+r.doc_wt
if d_sw not in bins.keys():
bins[d_sw] = []
bins[d_sw].append(r)
with open('all_results.obj', 'wb') as w_obj:
pickle.dump(all_results, w_obj)
with open('all_bins.obj', 'wb') as w_obj:
pickle.dump(bins, w_obj)
else:
print('reading results')
with open('all_results.obj', 'rb') as r_obj:
all_results = pickle.load(r_obj)
with open('all_bins.obj', 'rb') as r_obj:
bins = pickle.load(r_obj)
print('found %i bins' % len(bins))
if not os.path.isfile('graph_%i_%i.obj' % (args['diff_by'], args['doc_diff_by'])):
print('creating graph')
G = nx.Graph()
[G.add_node(a) for a in bins.keys()]
print('found %i nodes' % G.number_of_nodes())
for n1 in G.nodes_iter():
for n2 in G.nodes_iter():
if n1 != n2:
coh_sw_1, coh_sw_2 = n1.split('-')[0], n2.split('-')[0]
doc_sw_1, doc_sw_2 = n1.split('-')[1], n2.split('-')[1]
doc_wt_1, doc_wt_2 = n1.split('-')[2], n2.split('-')[2]
doc_diff = 1 if are_docs_from_diff_clusters(doc_wt_1, doc_wt_2) else 0
symm_switch = switch_symm_changer(doc_sw_2)
if switches_differ({'diff_by': args['diff_by']}, coh_sw_1, coh_sw_2) >= args['diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, doc_sw_2) + doc_diff >= args['doc_diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, symm_switch) + doc_diff >= args['doc_diff_by']:
G.add_edge(n1, n2)
with open('graph_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']), 'wb') as w_obj:
pickle.dump(G, w_obj)
else:
print('reading graph')
with open('graph_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']), 'rb') as r_obj:
G = pickle.load(r_obj)
if not os.path.isfile('max_cliques_%i_%i.obj' % (args['diff_by'], args['doc_diff_by'])):
cliques = [a for a in nx.find_cliques(G)]
max_len = max([len(a) for a in cliques])
max_cliques = [a for a in cliques if len(a) == max_len]
print('there are %i cliques with %i structures in each for diff_by=%i doc_diff_by=%i' %
(len(max_cliques), max_len, args['diff_by'], args['doc_diff_by']))
with open('max_cliques_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']), 'wb') as w_obj:
pickle.dump(max_cliques, w_obj)
else:
print('reading cliques')
with open('max_cliques_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']), 'rb') as r_obj:
max_cliques = pickle.load(r_obj)
occurences = {a.name: 0 for clq in max_cliques for k in clq for a in bins[k]}
for clq in max_cliques:
print('in clq', clq)
for k in clq:
# print(bins[k][0].name)
print('\n'.join(set([a.name for a in bins[k]])))
for a in bins[k]:
occurences[a.name] += 1
for w in sorted(occurences, key=occurences.get, reverse=False):
print(w, occurences[w])
parser = argparse.ArgumentParser()
parser.add_argument('-score_file')
parser.add_argument('-coh_seqs_file')
parser.add_argument('-doc_seqs_file')
parser.add_argument('-mode')
parser.add_argument('-n', type=int, default=1)
parser.add_argument('-diff_by', type=int, default=2)
parser.add_argument('-doc_diff_by', type=int, default=1)
parser.add_argument('-score_dir', type=str, default='./')
parser.add_argument('-purples_threshold', type=int, default=50)
args = vars(parser.parse_args())
if args['mode'] != 'bins_diagonal' and args['mode'] != 'post_pred_cliques' and args['mode'] != 'minidiagonal_cliques':
scores = score2dict(args['score_file'])
run_filters = ''#generate_run_filters()
coh_seq_dict = read_multi_fastas(args['coh_seqs_file'], suffix_to_remove='.pdb.gz')
doc_seq_dict = read_multi_fastas(args['doc_seqs_file'], suffix_to_remove='.pdb.gz')
if args['mode'] == 'switches_n_cliques':
switches, num_bins = make_switches(args, scores, run_filters, coh_seq_dict)
max_cliques = best_cliques(args, list(switches.keys()))
with open('switches.obj', 'wb') as sw_file:
pickle.dump(switches, sw_file)
with open('max_cliques.obj', 'wb') as clq_file:
pickle.dump(max_cliques, clq_file)
elif args['mode'] == 'choose_by_identity':
with open('switches.obj', 'rb') as sw_in:
switches = pickle.load(sw_in)
with open('max_cliques.obj', 'rb') as clq_file:
def prepare_data(in_df: pd.DataFrame) -> (pd.DataFrame, pd.DataFrame):
"""
:rtype: (pd.DataFrame, pd.DataFrame)
"""
rachel_root = '/home/labs/fleishman/jonathaw/decision_tree/'
cohs_non_aln = read_multi_fastas(rachel_root+'cohesins_from_rachel_and_vered.fasta', suffix_to_remove='/', lower=True)
docs_non_aln = read_multi_fastas(rachel_root+'dockerins_from_rachel_and_vered.fasta', suffix_to_remove='/', lower=True)
cohs = read_multi_fastas(rachel_root+'cohesins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/', lower=True)
docs = read_multi_fastas(rachel_root+'dockerins_from_rachel_and_vered.fasta_aln', suffix_to_remove='/', lower=True)
interface_positions = parse_interface_positions()
coh_poses = {coh: {typ: cohs[coh].non_aligned_position_at_aligned(pos) for typ, pos in typos.items()} for coh, typos
in interface_positions['coh'].items()}
doc_poses = {doc: {typ: docs[doc].non_aligned_position_at_aligned(pos) for typ, pos in typos.items()} for doc, typos
in interface_positions['doc'].items()}
validate_aligned_non_aligned_interface_positions(interface_positions['coh'], cohs, cohs_non_aln)
validate_aligned_non_aligned_interface_positions(interface_positions['doc'], docs, docs_non_aln)
coh_crys_seqs = [c for c in cohs.values() if c.name in ['1ohz', '2b59', '2ozn', '2vn5', '2y3n', '3ul4',
'4fl4', '4fl5', '4dh2', '4uyp', '5new']]
doc_crys_seqs = [d for d in docs.values() if d.name in ['1ohz', '2b59', '2ozn', '2vn5', '2y3n', '3ul4',
'4fl4', '4fl5', '4dh2', '4uyp', '5new']]
# columns = ['coh_name', 'doc_name', 'coh_seq', 'doc_seq'] + \
# ['%s_%s' % (typ, aa) for typ in ordered_positions['coh'] if 'core' in typ for aa in aas] + \
# ['%s_%s' % (typ, aa) for typ in ordered_positions['coh'] if 'rim' in typ for aa in types] + \
# ['%s_%s' % (typ, aa) for typ in ordered_positions['doc'] if 'core' in typ for aa in aas] + \
# ['%s_%s' % (typ, aa) for typ in ordered_positions['doc'] if 'rim' in typ for aa in types] + ['binders']
out_df = pd.DataFrame(index=range(1, len(in_df.index)), columns=columns)
id_columns = ['coh_name', 'doc_name', 'coh_seq', 'doc_seq'] + ordered_positions['coh'] + ordered_positions['doc']
identities_df = pd.DataFrame(index=range(1, len(in_df.index)), columns=id_columns)
for i in range(1, len(in_df.index)+1):
# find which crystal coh+doc are most similar
similar_coh, coh_iden = highest_seq_similarity(coh_crys_seqs, in_df.loc[i]['coh_seq'])
similar_doc, doc_iden = highest_seq_similarity(doc_crys_seqs, in_df.loc[i]['doc_seq'])
# get aligned positions accrotding to interface_positions
coh_identities = {typ: in_df.loc[i]['coh_seq'].get_aligned_positions([pos])[0] for typ, pos in
coh_poses[similar_coh.name].items()}
doc_identities = {typ: in_df.loc[i]['doc_seq'].get_aligned_positions([pos])[0] for typ, pos in
doc_poses[similar_doc.name].items()}
# coh_ = in_df.loc[i]['coh_seq'].get_aligned_positions(coh_poses[similar_coh])
# doc_ = in_df.loc[i]['doc_seq'].get_aligned_positions(doc_poses[similar_doc])
coh_core = [core_res_to_identity(coh_identities[v], 'coh') for v in ordered_positions['coh'] if 'core' in v]
coh_rim = [rim_res_to_type_binary(coh_identities[v]) for v in ordered_positions['coh'] if 'rim' in v]
doc_core = [core_res_to_identity(doc_identities[v], 'doc') for v in ordered_positions['doc'] if 'core' in v]
doc_rim = [rim_res_to_type_binary(doc_identities[v]) for v in ordered_positions['doc'] if 'rim' in v]
coh_core_list, coh_rim_list = [], []
[coh_core_list.append(a) for b in coh_core for a in b]
[coh_rim_list.append(a) for b in coh_rim for a in b]
doc_core_list, doc_rim_list = [], []
[doc_core_list.append(a) for b in doc_core for a in b]
[doc_rim_list.append(a) for b in doc_rim for a in b]
out_df.loc[i] = [in_df.loc[i]['coh_name'], in_df.loc[i]['doc_name'], 0, 0] + coh_core_list + coh_rim_list + \
doc_core_list + doc_rim_list + [1 if in_df.loc[i]['binders'] else 0]
identities_df.loc[i] = [in_df.loc[i]['coh_name'], in_df.loc[i]['doc_name'], 0, 0] + \
[coh_identities[v] for v in ordered_positions['coh'] if 'core' in v] + \
[coh_identities[v] for v in ordered_positions['coh'] if 'rim' in v] + \
[doc_identities[v] for v in ordered_positions['doc'] if 'core' in v] + \
[doc_identities[v] for v in ordered_positions['doc'] if 'rim' in v]
return out_df, identities_df
def post_pred_cliques(args):
run_filters = generate_run_filters(args={
'ddg': 25.0,
'sasa': 1400,
'shape': 0.6,
'packstat': 0.6,
'buried_2': 3
})
if not os.path.isfile('./all_data.obj'):
sc_files = [a for a in os.listdir('./') if '.score' in a]
cohs_seqs = read_multi_fastas(
'/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/all_cohs.fasta'
)
docs_seqs = read_multi_fastas(
'/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/all_docs.fasta'
)
results = []
for sc_file in sc_files:
seq_name = '_'.join(sc_file.split('_')[1:8])
coh_name = seq_name + '.pdb.gz.A'
doc_name = seq_name + '.pdb.gz.B'
sc_dict = score2dict(sc_file)
ynum = re.search('y[0-9]{3}', sc_file).group(0)
passed, failed = all_who_pass_run_filters(args, sc_dict,
run_filters)
if len(passed) >= args['purples_threshold']:
r = Result(seq_name, cohs_seqs[coh_name], docs_seqs[doc_name],
len(passed))
results.append(r)
with open('./all_data.obj', 'wb') as fout:
pickle.dump(results, fout)
else:
with open('./all_data.obj', 'rb') as fin:
results = pickle.load(fin)
if not os.path.isfile('./graph.obj'):
result_dict = {i + 1: r for i, r in enumerate(results)}
G = nx.Graph()
[G.add_node(a) for a in result_dict.keys()]
for n1 in G.nodes_iter():
for n2 in G.nodes_iter():
if n1 != n2:
coh_sw_1, coh_sw_2 = result_dict[
n1].coh_switch, result_dict[n2].coh_switch
doc_sw_1, doc_sw_2 = result_dict[
n1].doc_switch, result_dict[n2].doc_switch
doc_wt_1, doc_wt_2 = result_dict[n1].doc_wt, result_dict[
n2].doc_wt
doc_diff = 1 if are_docs_from_diff_clusters(
doc_wt_1, doc_wt_2) else 0
symm_switch = switch_symm_changer(doc_sw_2)
if switches_differ({'diff_by': args['diff_by']}, coh_sw_1, coh_sw_2) >= args['diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, doc_sw_2) + doc_diff >= args['doc_diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, symm_switch) + doc_diff >= args['doc_diff_by']:
G.add_edge(n1, n2)
print('adding edge\n', result_dict[n1], '\n',
result_dict[n2])
else:
print('NOT\n', result_dict[n1], '\n', result_dict[n2])
cliques = [a for a in nx.find_cliques(G)]
max_len = max([len(a) for a in cliques])
max_cliques = [a for a in cliques if len(a) == max_len]
for clq in max_cliques:
print(clq, '\n', '\n'.join([str(result_dict[a]) for a in clq]))
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-mode')
parser.add_argument('-make_dt', type=bool, default=False)
parser.add_argument('-coh_name', type=str)
args = vars(parser.parse_args())
if args['make_dt']:
print('making decision tree')
data_df = parse_binding_data()
prepared_df, identities_df = prepare_data(data_df)
decision_tree, features = create_decision_tree(prepared_df)
pickle.dump(decision_tree, open(decision_tree_root+'decision_tree_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
pickle.dump(features, open(decision_tree_root+'features_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
pickle.dump(prepared_df, open(decision_tree_root+'prepared_df_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
pickle.dump(identities_df, open(decision_tree_root+'identities_df_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
else:
print('reading decision tree')
decision_tree = pickle.load(open(decision_tree_root+'decision_tree_%s.obj' % time_to_use, 'rb'))
features = pickle.load(open(decision_tree_root+'features_%s.obj' % time_to_use, 'rb'))
prepared_df = pickle.load(open(decision_tree_root+'prepared_df_%s.obj' % time_to_use, 'rb'))
identities_df = pickle.load(open(decision_tree_root+'identities_df_%s.obj' % time_to_use, 'rb'))
if args['mode'] == 'k_fold':
k_fold_test(prepared_df)
elif args['mode'] == 'validate_df':
validate_data_frame(data_df, prepared_df)
elif args['mode'] == 'analyse_identities_df':
analyse_identity_df(identities_df)
elif args['mode'] == 'create_dt':
with open('decision_tree.dot', 'w') as fout:
print('creating decision tree.dot')
export_graphviz(decision_tree, out_file=fout, feature_names=features)
compare_observed_to_predicted(decision_tree, data_df, prepared_df[features])
elif args['mode'] == 'follow':
seq_to_follow(prepared_df, '2b59', '2b59')
elif args['mode'] == 'predict_all_designs_diagonal':
print('getting design sequences')
design_df = get_design_data()
print('making prediction')
design_df['predict'] = decision_tree.predict(design_df[features])
with open(design_data_root+'diagonal_prediciton.txt', 'w+') as fout:
pd.set_option('display.max_rows', 9999999999999999999)
fout.write(str(design_df.loc[design_df['predict'] == 1]['coh_name']))
elif args['mode'] == 'pickle_design_sequences':
dsn_cohs = read_multi_fastas(design_data_root+'all_designed_cohs.fasta', suffix_to_remove='.')
dsn_docs = read_multi_fastas(design_data_root+'all_designed_docs.fasta', suffix_to_remove='.')
pickle.dump(dsn_cohs, open(design_data_root+'dsn_cohs_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
pickle.dump(dsn_docs, open(design_data_root+'dsn_docs_%s.obj' % time.strftime("%d.%0-m"), 'wb'))
elif args['mode'] == 'predict_by_coh':
design_df = get_design_data_coh_vs_all(args['coh_name'])
print('predicting!!')
design_df['predict'] = decision_tree.predict(design_df[features])
with open(design_data_root+'all_vs_all_decision_tree_6Jan/'
+args['coh_name']+'.txt', 'w+') as fout:
pd.set_option('display.max_rows', len(design_df))
fout.write(str(design_df[[0, 1, -1]])+'\n')
pd.reset_option('display.max_rows')
else:
print('no mode found')
79,
81,
82,
83,
85,
87,
115,
116,
118,
119,
121,
123,
125,
127,
]
fastas = read_multi_fastas("/home/labs/fleishman/jonathaw/data/pssm/cohs/making/1ohz_passed_thresholds.fasta")
positives, negatives, neutrals, totals = [], [], [], []
for k, v in fastas.items():
charge_config = extract_charge_configuration(v, positions)
positives.append(charge_config.count("p"))
negatives.append(charge_config.count("n"))
neutrals.append(charge_config.count("c"))
totals.append(charge_config.count("p") + charge_config.count("n"))
if k == "1ohz":
print("at 1ohz found %i %s" % (charge_config.count("p"), "positives"))
print("at 1ohz found %i %s" % (charge_config.count("n"), "negatives"))
print("at 1ohz found %i %s" % (charge_config.count("c"), "neutrals"))
print("at 1ohz found %i %s" % (charge_config.count("p") + charge_config.count("n"), "totals"))
bins = range(max(positives + negatives + neutrals) + 1)
plt.hist(positives, bins=bins, color="b", label="positives")
plt.hist(negatives, bins=bins, color="r", label="negatives")
def extract_charge_configuration(seq: AASeq, positions: list):
res_in_poses = seq.get_positions(positions)
charge = [
res2charge[a] if a in res2charge.keys() else 'c' for a in res_in_poses
]
return charge
if __name__ == '__main__':
positions = [
32, 33, 35, 37, 39, 63, 66, 68, 70, 73, 75, 77, 79, 81, 82, 83, 85, 87,
115, 116, 118, 119, 121, 123, 125, 127
]
fastas = read_multi_fastas(
'/home/labs/fleishman/jonathaw/data/pssm/cohs/making/1ohz_passed_thresholds.fasta'
)
positives, negatives, neutrals, totals = [], [], [], []
for k, v in fastas.items():
charge_config = extract_charge_configuration(v, positions)
positives.append(charge_config.count('p'))
negatives.append(charge_config.count('n'))
neutrals.append(charge_config.count('c'))
totals.append(charge_config.count('p') + charge_config.count('n'))
if k == '1ohz':
print('at 1ohz found %i %s' %
(charge_config.count('p'), 'positives'))
print('at 1ohz found %i %s' %
(charge_config.count('n'), 'negatives'))
print('at 1ohz found %i %s' %
(charge_config.count('c'), 'neutrals'))
def minidiagonal_cliques(args):
original_names = parse_name_translation(
'/home/labs/fleishman/jonathaw/no_backup/designs/multi_docs_15Oct/recliques_4Nov/clique_6_pdbs/mini_diagonal_11Nov/minidiagonal_pdbs/translate_names.txt'
)
coh_seqs = read_multi_fastas(args['coh_seqs_file'], suffix_to_remove='.A')
doc_seqs = read_multi_fastas(args['doc_seqs_file'], suffix_to_remove='.B')
if not os.path.isfile('all_results.obj') or not os.path.isfile(
'all_bins.obj'):
print('creating bins and results')
all_results, bins = {}, {}
for design in coh_seqs.keys():
r = Result(design,
coh_seqs[design],
doc_seqs[design],
0,
j=True,
originals=original_names)
if not 2 < r.coh_switch.count('n') + r.coh_switch.count(
'p') < 8 and not 2 <= r.coh_switch.count('p') <= 3:
continue
all_results[design] = r
d_sw = r.coh_switch + '-' + r.doc_switch + '-' + r.doc_wt
if d_sw not in bins.keys():
bins[d_sw] = []
bins[d_sw].append(r)
with open('all_results.obj', 'wb') as w_obj:
pickle.dump(all_results, w_obj)
with open('all_bins.obj', 'wb') as w_obj:
pickle.dump(bins, w_obj)
else:
print('reading results')
with open('all_results.obj', 'rb') as r_obj:
all_results = pickle.load(r_obj)
with open('all_bins.obj', 'rb') as r_obj:
bins = pickle.load(r_obj)
print('found %i bins' % len(bins))
if not os.path.isfile('graph_%i_%i.obj' %
(args['diff_by'], args['doc_diff_by'])):
print('creating graph')
G = nx.Graph()
[G.add_node(a) for a in bins.keys()]
print('found %i nodes' % G.number_of_nodes())
for n1 in G.nodes_iter():
for n2 in G.nodes_iter():
if n1 != n2:
coh_sw_1, coh_sw_2 = n1.split('-')[0], n2.split('-')[0]
doc_sw_1, doc_sw_2 = n1.split('-')[1], n2.split('-')[1]
doc_wt_1, doc_wt_2 = n1.split('-')[2], n2.split('-')[2]
doc_diff = 1 if are_docs_from_diff_clusters(
doc_wt_1, doc_wt_2) else 0
symm_switch = switch_symm_changer(doc_sw_2)
if switches_differ({'diff_by': args['diff_by']}, coh_sw_1, coh_sw_2) >= args['diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, doc_sw_2) + doc_diff >= args['doc_diff_by'] and \
switches_differ({'diff_by': args['doc_diff_by']}, doc_sw_1, symm_switch) + doc_diff >= args['doc_diff_by']:
G.add_edge(n1, n2)
with open('graph_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']),
'wb') as w_obj:
pickle.dump(G, w_obj)
else:
print('reading graph')
with open('graph_%i_%i.obj' % (args['diff_by'], args['doc_diff_by']),
'rb') as r_obj:
G = pickle.load(r_obj)
if not os.path.isfile('max_cliques_%i_%i.obj' %
(args['diff_by'], args['doc_diff_by'])):
cliques = [a for a in nx.find_cliques(G)]
max_len = max([len(a) for a in cliques])
max_cliques = [a for a in cliques if len(a) == max_len]
print(
'there are %i cliques with %i structures in each for diff_by=%i doc_diff_by=%i'
%
(len(max_cliques), max_len, args['diff_by'], args['doc_diff_by']))
with open(
'max_cliques_%i_%i.obj' %
(args['diff_by'], args['doc_diff_by']), 'wb') as w_obj:
pickle.dump(max_cliques, w_obj)
else:
print('reading cliques')
with open(
'max_cliques_%i_%i.obj' %
(args['diff_by'], args['doc_diff_by']), 'rb') as r_obj:
max_cliques = pickle.load(r_obj)
occurences = {
a.name: 0
for clq in max_cliques for k in clq for a in bins[k]
}
for clq in max_cliques:
print('in clq', clq)
for k in clq:
# print(bins[k][0].name)
print('\n'.join(set([a.name for a in bins[k]])))
for a in bins[k]:
occurences[a.name] += 1
for w in sorted(occurences, key=occurences.get, reverse=False):
print(w, occurences[w])
