def summarize_residues(mutations,
pdb_info,
radius,
rASA,
dssp,
tmp_dir,
quiet=True):
# iterate over each structure
logger.info('Running of PDB structures . . .')
output = [[
'structure', 'tumor type', '# buried residues',
'# protein interface residues', '# nucleic acid interface residues',
'total residues', '# buried mutations',
'# protien interface mutations', '# nucleic acid interface mutations',
'total # mutations', 'burial p-value', 'protein interface p-value',
'nucleic acid interface p-value'
]]
for structure_id in pdb_info:
#if structure_id.startswith('ENSP') or structure_id.startswith('NP_'):
#continue
#print structure_id
# get pdb info
struct_info = pdb_info[structure_id]
pdb_path = struct_info.pop('path')
# read in structure
structure = utils.read_structure(pdb_path, structure_id, quiet=quiet)
if structure is None:
continue
# make a list of all chain letters in structure
struct_chains = []
for k in struct_info.keys():
struct_chains.extend(struct_info[k])
structure_mutations = mutations.get(structure_id, [])
# skip structure if no mutations
if not structure_mutations:
continue
# separate out mutation info
ttypes, mres, mcount, mchains = zip(
*structure_mutations) # if model_mutations else ([], [], [])
# stratify mutations by their tumor type
# ttype_ixs is a dictionary that contains
# ttype as the keys and a list of relevant
# indices as the values
unique_ttypes = set(ttypes)
ttype_ixs = {
t: [i for i in range(len(mcount)) if ttypes[i] == t]
for t in unique_ttypes
}
#ttype_ixs['PANCAN'] = range(len(mcount))
# add PANCAN as a "tumour type"
unique_ttypes = list(unique_ttypes)
#unique_ttypes.append('PANCAN')
# obtain relevant info from structure
tmp_info = pstruct.get_structure_info(structure, mchains, mres, mcount,
struct_chains, ttype_ixs)
(mut_res_centers_of_geometry, mut_res_mutation_counts,
all_res_centers_of_geometry, models) = tmp_info
annotated_chains = {
chain
for description in struct_info
for chain in struct_info[description]
}
# find buried residues
buried_res = pstruct.get_buried_residues(structure, rASA, tmp_dir,
dssp)
tmp_buried = [
res_id for res_id in buried_res if res_id[2] in annotated_chains
]
total_res = len(tmp_buried)
buried_res_info = {(info[1], info[2], info[3])
for info in tmp_buried if info[-1] == 1}
num_buried_res = len(buried_res_info)
# find interface residues for proteins and nucleic acids
interface_res = pstruct.get_interface_residues(structure, radius)
interface_prot_info = {(res_id[1], res_id[2], res_id[3][1])
for res_id in interface_res
if (res_id[2] in annotated_chains)
and interface_res[res_id][0] == 1}
interface_na_info = {(res_id[1], res_id[2], res_id[3][1])
for res_id in interface_res
if (res_id[2] in annotated_chains)
and sum(interface_res[res_id][1:]) >= 1}
num_interface_prot_res = len(interface_prot_info)
num_interface_na_res = len(interface_na_info)
# iterate through each tumour type
pan_counts = []
pan_buried_counts = []
pan_interface_prot_counts, pan_interface_na_counts = [], []
tmp_output = []
for tumour in unique_ttypes:
# skip tumor types if not one specified
#if (not opts['tumor_type'] == tumour and not opts['tumor_type'] == 'EVERY'):
#continue
# draw information for the specific tumour type
t_mut_res_centers_of_geometry = mut_res_centers_of_geometry[tumour]
t_mut_res_mutation_counts = mut_res_mutation_counts[tumour]
# count total mutations in structure while
# avoiding double counting due to same id and chain
# being on multiple models
obs_models = []
obs_chains = []
total_mutations = 0
total_buried_muts = 0
total_interface_prot_muts, total_interface_na_muts = 0, 0
banned_chains = set()
#if not tumour == 'PANCAN':
if True:
for k in t_mut_res_mutation_counts:
mutations_to_add = t_mut_res_mutation_counts[k]
# prevent double counting
cur_model = k[1]
cur_chain = k[2]
cur_pos = k[3][1]
for i in range(len(obs_models)):
if not cur_model == obs_models[
i] and cur_chain == obs_chains[i]:
mutations_to_add = 0
break
if (cur_chain, cur_pos) in banned_chains:
mutations_to_add = 0
# add all equivalent chains to banned list
equiv_chains = pstruct.find_eq_letters(
struct_info, cur_chain)
if equiv_chains is not None:
equiv_pos = set([(e, cur_pos) for e in equiv_chains])
banned_chains |= equiv_pos - set([(cur_chain, cur_pos)
])
# add to total mutation count
total_mutations += mutations_to_add
# current residue of interest
curr_res = (cur_model, cur_chain, cur_pos)
# add buried residue mutation counts
is_buried = [(m, c[0], c[1]) in buried_res_info
for c in equiv_pos for m in range(4)]
#if (curr_res in buried_res_info):
if any(is_buried):
total_buried_muts += mutations_to_add
pan_buried_counts.append(mutations_to_add)
# add interface residue mutation counts
is_interface_prot = [(m, c[0], c[1]) in interface_prot_info
for c in equiv_pos for m in range(4)]
is_interface_na = [(m, c[0], c[1]) in interface_na_info
for c in equiv_pos for m in range(4)]
#if (curr_res in interface_info):
if any(is_interface_prot):
total_interface_prot_muts += mutations_to_add
pan_interface_prot_counts.append(mutations_to_add)
if any(is_interface_na):
total_interface_na_muts += mutations_to_add
pan_interface_na_counts.append(mutations_to_add)
# mark chains/models
obs_models.append(k[1])
obs_chains.append(k[2])
pan_counts.append(total_mutations)
else:
total_mutations = sum(pan_counts)
total_buried_muts = sum(pan_buried_counts)
total_interface_prot_muts = sum(pan_interface_prot_counts)
total_interface_na_muts = sum(pan_interface_na_counts)
tmp_output.append([
structure_id,
tumour,
num_buried_res,
num_interface_prot_res,
num_interface_na_res,
total_res,
total_buried_muts,
total_interface_prot_muts,
total_interface_na_muts,
total_mutations,
])
output.extend(tmp_output)
return output
def main(opts):
# read in the PDB info file
pdb_info = utils.read_pdb_info(opts['pdb_info'])
# read in multiple testing file
mtc = read_delim(opts['multiple_testing'])
header = mtc.pop(0)
ttype_ix = header.index('Tumor Type')
qval_ix = header.index('q-value')
gene_ix = header.index('HUGO Symbol')
tx_ix = header.index('Sequence Ontology Transcript')
res_ix = header.index('CRAVAT Res')
#mtc.sort(key=lambda x: x[0])
# iterate through each tumor type
output = []
gene2graph_all = {} # graphs for combined tumor types
uniq_ttypes = set(m[ttype_ix] for m in mtc)
for ttype in uniq_ttypes:
logger.info('Working on {0} . . .'.format(ttype))
# initialize the graph to empty
gene2graph = {} # graph for an individual tumor type
# get the significant residues for the tumor type
mtc_ttype = [m for m in mtc
if (m[ttype_ix] == ttype) and (float(m[qval_ix])<=opts['q_value'])]
significant_res = set([(m[gene_ix], m[tx_ix], int(m[res_ix]))
for m in mtc_ttype])
# read annotation file
annotation_file = os.path.join(opts['annotation_dir'], 'mupit_mutations_' + ttype)
annotation, col_pos = read_mupit_file(annotation_file, significant_res)
pdb_ix = col_pos['pdb']
anot_gene_ix = col_pos['gene']
anot_tx_ix = col_pos['tx']
anot_res_ix = col_pos['res']
# sort by structure
annotation.sort(key=lambda x: x[pdb_ix])
for pdb_id, grp in it.groupby(annotation, lambda x: x[pdb_ix]):
struct_info = pdb_info[pdb_id].copy()
pdb_path = struct_info.pop('path')
struct_chains = []
for d in struct_info:
struct_chains.extend(struct_info[d])
#pdb_path = pdb2path[pdb_id]
struct = utils.read_structure(pdb_path, pdb_id)
if struct is None:
continue # skip if pdb file not found
# calculate the centers of geometry
cog = pstruct.calc_center_of_geometry(struct, struct_chains)
# contains relevant mupit annotations for this pdb
tmp = list(grp)
# get significant residues
signif_struct_info = {}
for s in tmp:
try:
tmp_pos = (s[col_pos['chain']], int(s[col_pos['pdb_res']]))
except:
print 'int error'
continue
signif_struct_info[tmp_pos] = (s[anot_gene_ix], s[anot_tx_ix], s[anot_res_ix])
# update the graph to reflect info from the current structure
gene2graph = update_graph(gene2graph, cog, signif_struct_info,
struct, opts['radius'])
# update graph for the combined cross-tumor type regions
banned_ttypes = ['COAD', 'READ', 'PANCAN12', 'CHOL', 'SARC',
'TGCT', 'THYM', 'UVM']
if ttype not in banned_ttypes:
gene2graph_all = update_graph(gene2graph_all, cog, signif_struct_info,
struct, opts['radius'])
# format the results into the output list
tmp_out = retrieve_components(gene2graph, ttype)
output += tmp_out
logger.info('Finished {0}'.format(ttype))
# update output to contain cross-tumor type reference regions
tmp_out = retrieve_components(gene2graph_all, 'REF')
output += tmp_out
# write output
with open(opts['output'], 'wb') as handle:
for line in output:
handle.write('\t'.join(line)+'\n')
logger.info('Finished Successfully!!!')
def main(opts):
# read in the PDB info file
pdb_info = utils.read_pdb_info(opts['pdb_info'])
# read in multiple testing file
mtc = read_delim(opts['multiple_testing'])
header = mtc.pop(0)
ttype_ix = header.index('Tumor Type')
qval_ix = header.index('q-value')
gene_ix = header.index('HUGO Symbol')
tx_ix = header.index('Sequence Ontology Transcript')
res_ix = header.index('CRAVAT Res')
#mtc.sort(key=lambda x: x[0])
# iterate through each tumor type
output = []
gene2graph_all = {} # graphs for combined tumor types
uniq_ttypes = set(m[ttype_ix] for m in mtc)
for ttype in uniq_ttypes:
logger.info('Working on {0} . . .'.format(ttype))
# initialize the graph to empty
gene2graph = {} # graph for an individual tumor type
# get the significant residues for the tumor type
mtc_ttype = [
m for m in mtc if (m[ttype_ix] == ttype) and (
float(m[qval_ix]) <= opts['q_value'])
]
significant_res = set([(m[gene_ix], m[tx_ix], int(m[res_ix]))
for m in mtc_ttype])
# read annotation file
annotation_file = os.path.join(opts['annotation_dir'],
'mupit_mutations_' + ttype)
annotation, col_pos = read_mupit_file(annotation_file, significant_res)
pdb_ix = col_pos['pdb']
anot_gene_ix = col_pos['gene']
anot_tx_ix = col_pos['tx']
anot_res_ix = col_pos['res']
# sort by structure
annotation.sort(key=lambda x: x[pdb_ix])
for pdb_id, grp in it.groupby(annotation, lambda x: x[pdb_ix]):
# fringe case
if pdb_id not in pdb_info:
print('skipping ' + pdb_id)
continue
# get path info
struct_info = pdb_info[pdb_id].copy()
pdb_path = struct_info.pop('path')
struct_chains = []
for d in struct_info:
struct_chains.extend(struct_info[d])
#pdb_path = pdb2path[pdb_id]
struct = utils.read_structure(pdb_path, pdb_id)
if struct is None:
continue # skip if pdb file not found
# calculate the centers of geometry
cog = pstruct.calc_center_of_geometry(struct, struct_chains)
# contains relevant mupit annotations for this pdb
tmp = list(grp)
# get significant residues
signif_struct_info = {}
for s in tmp:
try:
tmp_pos = (s[col_pos['chain']], int(s[col_pos['pdb_res']]))
except:
print 'int error'
continue
signif_struct_info[tmp_pos] = (s[anot_gene_ix], s[anot_tx_ix],
s[anot_res_ix])
# update the graph to reflect info from the current structure
gene2graph = update_graph(gene2graph, cog, signif_struct_info,
struct, opts['radius'])
# update graph for the combined cross-tumor type regions
gene2graph_all = update_graph(gene2graph_all, cog,
signif_struct_info, struct,
opts['radius'])
# format the results into the output list
tmp_out = retrieve_components(gene2graph, ttype)
output += tmp_out
logger.info('Finished {0}'.format(ttype))
# update output to contain cross-tumor type reference regions
tmp_out = retrieve_components(gene2graph_all, 'REF')
output += tmp_out
# write output
with open(opts['output'], 'wb') as handle:
for line in output:
handle.write('\t'.join(line) + '\n')
logger.info('Finished Successfully!!!')
def main(opts):
# read in the PDB info file
pdb_info = utils.read_pdb_info(opts['pdb_info'])
# use external module to separate out the residues in the hotspot.py output
# onto separate lines
mtc = read_residue_info(opts['input'])
pval_thresholds = read_thresholds(opts['significance'])
# read in multiple testing file
#mtc = read_delim(opts['multiple_testing'])
header = mtc.pop(0)
ttype_ix = header.index('Tumor Type')
struct_ix = header.index('Structure')
model_ix = header.index('Model')
chain_ix = header.index('Chain')
res_ix = header.index('Mutation Residues')
pval_ix = header.index('Hotspot P-value')
# iterate through each tumor type
output = []
uniq_ttypes = set(m[ttype_ix] for m in mtc)
for ttype in uniq_ttypes:
logger.info('Working on {0} . . .'.format(ttype))
# if there is no pval threshold, nothing is significant
if not ttype in pval_thresholds:
continue
# get the significant residues for the tumor type
mtc_ttype = [m for m in mtc
if (m[ttype_ix] == ttype) and (float(m[pval_ix])<=pval_thresholds[ttype])]
# ANY EQUIVALENT COPY THING FOR STRUCTURES?
# significant_res = set([(m[gene_ix], m[tx_ix], int(m[res_ix]))
# for m in mtc_ttype])
#significant_res = list(mtc_ttype)
significant_res = [(m[struct_ix], m[chain_ix], int(m[res_ix]))
for m in mtc_ttype]
# read annotation file
annotation_file = os.path.join(opts['annotation_dir'],
'mupit_mutations_' + ttype)
all_annotation, col_pos = read_mupit_file(annotation_file, significant_res)
pdb_ix = col_pos['pdb']
anot_gene_ix = col_pos['gene']
anot_tx_ix = col_pos['tx']
anot_res_ix = col_pos['res']
# sort by structure
all_annotation.sort(key=lambda x: x[pdb_ix])
for pdb_id, grp in it.groupby(all_annotation, lambda x: x[pdb_ix]):
# initialize the graph to empty
struct2graph = {}
struct_info = pdb_info[pdb_id].copy()
pdb_path = struct_info.pop('path')
struct_chains = []
for d in struct_info:
struct_chains.extend(struct_info[d])
#pdb_path = pdb2path[pdb_id]
struct = utils.read_structure(pdb_path, pdb_id)
if struct is None:
continue # skip if pdb file not found
# calculate the centers of geometry
all_cogs = pstruct.calc_center_of_geometry(struct, struct_chains)
# contains relevant mupit annotations for this pdb
tmp = list(grp)
# get significant residues
signif_struct_info = {}
non_signif_struct_info = {}
for s in tmp:
try:
tmp_pos = (s[col_pos['chain']], int(s[col_pos['pdb_res']]))
except:
continue
if (s[col_pos['pdb']], s[col_pos['chain']], int(s[col_pos['pdb_res']])) in significant_res:
signif_struct_info[tmp_pos] = (s[pdb_ix], s[anot_tx_ix], int(s[anot_res_ix]))
else:
non_signif_struct_info[tmp_pos] = (s[pdb_ix], s[anot_tx_ix], int(s[anot_res_ix]))
#print "Pushing update", pdb_id
# update the graph to reflect info from the current structure
struct2graph, signif_res_neighbours = update_graph(struct2graph, all_cogs, signif_struct_info, non_signif_struct_info,
struct, opts['radius'])
# format the results into the output list
tmp_out = retrieve_components(struct2graph, ttype, all_cogs, opts['radius'], signif_res_neighbours)
output += tmp_out
# format the results into the output list
# tmp_out = retrieve_components(struct2graph, ttype)
# output += tmp_out
logger.info('Finished {0}'.format(ttype))
# write output
with open(opts['output'], 'wb') as handle:
for line in output:
handle.write('\t'.join(line)+'\n')
logger.info('Finished Successfully!!!')
def main(opts):
"""Currently, performs analysis for the given genes. It attempts to use
any available PDB sturctures. It then loops through each protein chain
and tumor type.
"""
# read in data
logger.info('Reading in annotations . . .')
pdb_info = utils.read_pdb_info(opts['annotation'])
logger.info('Finished reading in annotations.')
logger.info('Reading in mutations . . .')
mutations = utils.read_mutations(opts['mutations'])
logger.info('Finished reading in mutations.')
# iterate over each structure
logger.info('Running of PDB structures . . .')
output = []
num_pdbs = 0
num_missing_pdbs = 0
missing_pdb_list = []
error_pdb_structs = []
quiet = True if opts[
'log_level'] != "DEBUG" else False # flag indicating pdb warnings
pdb_parser = PDBParser(QUIET=quiet) # parser for pdb files
for structure_id in pdb_info:
print(structure_id)
# get pdb info
struct_info = pdb_info[structure_id]
pdb_path = struct_info.pop('path')
# read in structure
structure = utils.read_structure(pdb_path, structure_id, quiet=quiet)
if structure is None:
continue
# make a list of all chain letters in structure
struct_chains = []
for k in struct_info.keys():
struct_chains.extend(struct_info[k])
# get mutation info
structure_mutations = mutations.get(structure_id, [])
# skip structure if no mutations
if not structure_mutations:
continue
# separate out mutation info
ttypes, mres, mcount, mchains = zip(
*structure_mutations) # if model_mutations else ([], [], [])
# stratify mutations by their tumor type
# ttype_ixs is a dictionary that contains
# ttype as the keys and a list of relevant
# indices as the values
unique_ttypes = set(ttypes)
ttype_ixs = {
t: [i for i in range(len(mcount)) if ttypes[i] == t]
for t in unique_ttypes
}
unique_ttypes = list(unique_ttypes)
# obtain relevant info from structure
tmp_info = get_structure_info(structure, mchains, mres, mcount,
struct_chains, ttype_ixs)
(mut_res_centers_of_geometry, mut_res_mutation_counts,
all_res_centers_of_geometry, models) = tmp_info
if not all_res_centers_of_geometry:
logger.error('No available center of geometries for {0}'.format(
structure_id))
continue
# get neigbours for all residues
neighbors = find_neighbors(all_res_centers_of_geometry, opts['radius'])
# iterate through each tumour type
for tumour in unique_ttypes:
# skip tumor types if not one specified
if (not opts['tumor_type'] == tumour
and not opts['tumor_type'] == 'EVERY'):
continue
# draw information for the specific tumour type
t_mut_res_centers_of_geometry = mut_res_centers_of_geometry[tumour]
t_mut_res_mutation_counts = mut_res_mutation_counts[tumour]
mut_density = src.mutations.mutation_density(
t_mut_res_mutation_counts, neighbors)
mut_vals = mut_density.values()
if mut_vals:
max_obs_dens = max(mut_density.values())
else:
max_obs_dens = 0
# generate null distribution
# count total mutations in structure while
# avoiding double counting due to same id and chain
# being on multiple models
obs_models = []
obs_chains = []
total_mutations = 0
for k in t_mut_res_mutation_counts:
mutations_to_add = t_mut_res_mutation_counts[k]
for i in range(len(obs_models)):
if not k[1] == obs_models[i] and k[2] == obs_chains[i]:
mutations_to_add = 0
break
total_mutations += mutations_to_add
obs_models.append(k[1])
obs_chains.append(k[2])
# generate empirical null distribution
sim_null_dist = sim.generate_null_dist(
structure_id, models, struct_info, all_res_centers_of_geometry,
total_mutations, opts['num_simulations'], opts['seed'],
neighbors, opts['stop_criterion'], max_obs_dens)
# get a list of lists format for compute p values function
mut_list = [[res_id, mut_density[res_id]]
for res_id in mut_density]
if not t_mut_res_mutation_counts:
print("here")
# aditional information about p-values
# for specific residues in a structure
# compute p-values for observed
obs_pvals, sim_cdf = sim.compute_pvals(mut_list, sim_null_dist)
output.append([
structure_id,
tumour,
','.join([str(o[0][1]) for o in mut_list]),
','.join([str(o[0][2]) for o in mut_list]),
','.join([str(o[0][3][1]) for o in mut_list]),
','.join(
[str(t_mut_res_mutation_counts[o[0]]) for o in mut_list]),
','.join([str(o[1]) for o in mut_list]),
','.join(map(str, obs_pvals)),
])
# write output to file
output = [[
'Structure',
'Tumor Type',
'Model',
'Chain',
'Mutation Residues',
'Residue Mutation Count',
'Mutation Density',
'Hotspot P-value',
]] + output
with open(opts['output'], 'w') as handle:
csv.writer(handle, delimiter='\t',
lineterminator='\n').writerows(output)
# if user specified to log failed reading of pdbs
if opts['error_pdb'] and error_pdb_structs:
with open(opts['error_pdb'], 'w') as handle:
for bad_pdb in error_pdb_structs:
handle.write(bad_pdb + '\n')
print("NUM_MODEL_DIFF: " + str(sim.NUM_MODEL_DIFF))
print("NUM_CHAIN_DIFF: " + str(sim.NUM_CHAIN_DIFF))
print("STRUCT_MODEL_DIFF: " + str(sim.STRUCT_MODEL_DIFF))
print("STRUCT_CHAIN_DIFF: " + str(sim.STRUCT_CHAIN_DIFF))
logger.info('Finished successfully!')
def main(opts):
"""Currently, performs analysis for the given genes. It attempts to use
any available PDB sturctures. It then loops through each protein chain
and tumor type.
"""
# read in data
logger.info('Reading in annotations . . .')
pdb_info = utils.read_pdb_info(opts['annotation'])
logger.info('Finished reading in annotations.')
logger.info('Reading in mutations . . .')
mutations = utils.read_mutations(opts['mutations'])
logger.info('Finished reading in mutations.')
# iterate over each structure
logger.info('Running of PDB structures . . .')
output = []
num_pdbs = 0
num_missing_pdbs = 0
missing_pdb_list = []
error_pdb_structs = []
quiet = True if opts['log_level'] != "DEBUG" else False # flag indicating pdb warnings
pdb_parser = PDBParser(QUIET=quiet) # parser for pdb files
for structure_id in pdb_info:
print (structure_id)
# get pdb info
struct_info = pdb_info[structure_id]
pdb_path = struct_info.pop('path')
# read in structure
structure = utils.read_structure(pdb_path, structure_id, quiet=quiet)
if structure is None:
continue
# make a list of all chain letters in structure
struct_chains = []
for k in struct_info.keys():
struct_chains.extend(struct_info[k])
# get mutation info
structure_mutations = mutations.get(structure_id, [])
# skip structure if no mutations
if not structure_mutations:
continue
# separate out mutation info
ttypes, mres, mcount, mchains = zip(*structure_mutations) # if model_mutations else ([], [], [])
# stratify mutations by their tumor type
# ttype_ixs is a dictionary that contains
# ttype as the keys and a list of relevant
# indices as the values
unique_ttypes = set(ttypes)
ttype_ixs = {t: [i for i in range(len(mcount)) if ttypes[i]==t]
for t in unique_ttypes}
unique_ttypes = list(unique_ttypes)
# obtain relevant info from structure
tmp_info = get_structure_info(structure, mchains, mres, mcount,
struct_chains, ttype_ixs)
(mut_res_centers_of_geometry,
mut_res_mutation_counts,
all_res_centers_of_geometry,
models) = tmp_info
if not all_res_centers_of_geometry:
logger.error('No available center of geometries for {0}'.format(structure_id))
continue
# get neigbours for all residues
neighbors = find_neighbors(all_res_centers_of_geometry, opts['radius'])
# iterate through each tumour type
for tumour in unique_ttypes:
# skip tumor types if not one specified
if (not opts['tumor_type'] == tumour and not opts['tumor_type'] == 'EVERY'):
continue
# draw information for the specific tumour type
t_mut_res_centers_of_geometry = mut_res_centers_of_geometry[tumour]
t_mut_res_mutation_counts = mut_res_mutation_counts[tumour]
mut_density = src.mutations.mutation_density(t_mut_res_mutation_counts,
neighbors)
mut_vals = mut_density.values()
if mut_vals:
max_obs_dens = max(mut_density.values())
else:
max_obs_dens =0
# generate null distribution
# count total mutations in structure while
# avoiding double counting due to same id and chain
# being on multiple models
obs_models = []
obs_chains = []
total_mutations = 0
for k in t_mut_res_mutation_counts:
mutations_to_add = t_mut_res_mutation_counts[k]
for i in range(len(obs_models)):
if not k[1] == obs_models[i] and k[2] == obs_chains[i]:
mutations_to_add = 0
break
total_mutations += mutations_to_add
obs_models.append(k[1])
obs_chains.append(k[2])
# generate empirical null distribution
sim_null_dist = sim.generate_null_dist(structure_id, models, struct_info,
all_res_centers_of_geometry,
total_mutations,
opts['num_simulations'],
opts['seed'],
neighbors,
opts['stop_criterion'],
max_obs_dens)
# get a list of lists format for compute p values function
mut_list = [[res_id, mut_density[res_id]] for res_id in mut_density]
if not t_mut_res_mutation_counts:
print("here")
# aditional information about p-values
# for specific residues in a structure
# compute p-values for observed
obs_pvals, sim_cdf = sim.compute_pvals(mut_list, sim_null_dist)
output.append([structure_id, tumour,
','.join([str(o[0][1]) for o in mut_list]),
','.join([str(o[0][2]) for o in mut_list]),
','.join([str(o[0][3][1]) for o in mut_list]),
','.join([str(t_mut_res_mutation_counts[o[0]])
for o in mut_list]),
','.join([str(o[1]) for o in mut_list]),
','.join(map(str, obs_pvals)),])
# write output to file
output = [['Structure', 'Tumor Type', 'Model', 'Chain', 'Mutation Residues',
'Residue Mutation Count', 'Mutation Density', 'Hotspot P-value',
]] + output
with open(opts['output'], 'w') as handle:
csv.writer(handle, delimiter='\t', lineterminator='\n').writerows(output)
# if user specified to log failed reading of pdbs
if opts['error_pdb'] and error_pdb_structs:
with open(opts['error_pdb'], 'w') as handle:
for bad_pdb in error_pdb_structs:
handle.write(bad_pdb+'\n')
print("NUM_MODEL_DIFF: " + str(sim.NUM_MODEL_DIFF))
print("NUM_CHAIN_DIFF: " + str(sim.NUM_CHAIN_DIFF))
print("STRUCT_MODEL_DIFF: " + str(sim.STRUCT_MODEL_DIFF))
print("STRUCT_CHAIN_DIFF: " + str(sim.STRUCT_CHAIN_DIFF))
logger.info('Finished successfully!')
def get_buried_residues(structure, cutoff, tmp_dir, dssp_path):
"""Finds buried residues by using relative solvent accessible surface area.
"""
# get structure id
structure_id = structure.id
all_letters = set(string.ascii_uppercase) | set(string.ascii_lowercase)
# flatten models into a single model due to limitations of DSSP
id_map = {}
for k, model in enumerate(structure):
if k == 0:
#used_letters = set(model.child_dict.keys())
used_letters = set()
for chain in model:
if chain.get_id() == ' ':
chain.id = 'A'
#for l in used_letters:
id_map[(model.id, chain.id)] = (model.id, chain.id)
used_letters.add(chain.id)
new_model = model.id
else:
for chain in model:
left_over = all_letters - used_letters
if not left_over:
# if run out of chain letters just return nothing
return []
new_letter = left_over.pop()
used_letters.add(new_letter)
old_letter = chain.id
chain.id = new_letter
id_map[(new_model, new_letter)] = (model.id, old_letter)
# add numbers if there is not more letters left
if not (all_letters - used_letters):
all_letters.update(
set(string.digits) | set(string.punctuation))
model.id = new_model
# save new structure to tmp dir
io = Bio.PDB.PDBIO()
io.set_structure(structure)
tmp_path = os.path.join(tmp_dir, structure_id + '.pdb')
io.save(tmp_path)
# read in tmp structure
tmp_structure = utils.read_structure(tmp_path, structure_id, quiet=True)
# find the solvent accessibility for residues
dssp_results = Bio.PDB.DSSP(tmp_structure[0], tmp_path, dssp=dssp_path)
# get bfactors for each amino acid residue
bfacs_missing = [
r for r in tmp_structure.get_residues()
if Bio.PDB.is_aa(r) and 'CA' not in r.child_dict
]
bfacs = [
r['CA'].get_bfactor() for r in tmp_structure.get_residues()
if Bio.PDB.is_aa(r) and 'CA' in r.child_dict
]
mean_bfac = np.mean(bfacs)
std_bfac = np.std(bfacs)
# format output
output = []
for result in dssp_results:
# skip if not an amino acid
if not Bio.PDB.is_aa(result[0]):
continue
# format the ID
full_id = result[0].get_full_id()
#if full_id[2] == ' ':
#full_id[2] = 'A'
try:
orig_model_chain = list(id_map[full_id[1:3]])
except:
print full_id, id_map
raise
# fix missing letter for homology models
if orig_model_chain[1] == ' ':
orig_model_chain[1] = 'A'
# record whether it was buried
if 'CA' in result[0].child_dict:
norm_bfactor = (result[0]['CA'].get_bfactor() -
mean_bfac) / std_bfac
else:
norm_bfactor = None
line = [structure_id] + orig_model_chain + [
result[0].id[1], result[3], norm_bfactor
]
if result[3] <= cutoff:
line.append(1)
else:
line.append(0)
output.append(line)
# delete tmp file
if os.path.exists(tmp_path): os.remove(tmp_path)
return output
