def main():
# reference interactome name
# options: HI-II-14, IntAct
ref_interactome_name = 'HI-II-14'
# similarity measure used to calculate GO similarity
sim_measure = 'SimGIC'
# root ontologies on which GO similarity was calculated
ont_root = [
'biological_process', 'molecular_function', 'cellular_component'
]
# root ontology labels used to label output files and figures
ont_abv = {
'biological_process': 'bp',
'molecular_function': 'mf',
'cellular_component': 'cc'
}
# structural interactome names
struc_name = {'HI-II-14': 'Y2H-SI', 'IntAct': 'IntAct-SI'}
# interactome names
interactome_names = [
'Random interactions', 'Reference interactome',
'Structural interactome'
]
# interactome colors
interactome_colors = ['limegreen', 'steelblue', 'orangered']
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# parent directory of processed data files
procDir = dataDir / 'processed'
# directory of processed data files specific to interactome
interactomeDir = procDir / ref_interactome_name
# figure directory
figDir = Path('../figures') / ref_interactome_name / 'gosim'
# input data files
gosimFiles = [
interactomeDir / 'gosim' / ('allPPI_gosim_%s_%s.pkl' %
(ont_abv[root], sim_measure))
for root in ont_root
]
# create output directories if not existing
if not figDir.exists():
os.makedirs(figDir)
allgosim = {}
for root, gosimFile in zip(ont_root, gosimFiles):
with open(gosimFile, 'rb') as f:
allgosim[root] = pickle.load(f)
means, errors = [], []
for interactome_name in interactome_names:
gosim = [
allgosim[root][interactome_name]["gosim"] for root in ont_root
]
means.append([np.mean(s) for s in gosim])
errors.append([sderror(s) for s in gosim])
multi_bar_plot(means,
errors=errors,
xlabels=[r.title().replace('_', '\n') for r in ont_root],
ylabel='Gene ontology similarity\nof interaction partners',
ylabels=[round(x, 1) for x in np.arange(0, 0.6, 0.1)],
colors=interactome_colors,
edgecolor='k',
ewidth=1.5,
barwidth=0.2,
bargap=0.03,
fontsize=18,
capsize=5,
msize=8,
leg=[
'Random pairs',
'%s reference interactome' % ref_interactome_name,
struc_name[ref_interactome_name]
],
show=showFigs,
figdir=figDir,
figname='interactome_gosim_%s' % sim_measure)
def main():
# reference interactome name
# options: HI-II-14, IntAct
interactome_name = 'HI-II-14'
# similarity measure to calculate GO similarity
# options: Resnik, Lin, Jiang-Conrath, SimGIC, SimUI, SimIC, SimRel, Dice, SimTO
# SimNTO, Jaccard, Czekanowski-Dice, Cosine, GSESAME, SimICND, SimICNP
sim_measure = 'SimGIC'
# mixing strategy for merging GO term semantic similarities
# options: max, avg, BMA (best match average)
mix_method = 'BMA'
# root ontology on which GO similarity is calculated
# options: biological_process, molecular_function, cellular_component
ont_root = 'biological_process'
# root ontology labels used to label output files and figures
ont_abv = {
'biological_process': 'bp',
'molecular_function': 'mf',
'cellular_component': 'cc'
}
# list of ontological relationships to ignore
ont_ignore = None
# list of evidence codes to ignore
ec_ignore = None
# number of random interactions
numRandPairs = 10000
# interactome colors
interactome_colors = ['limegreen', 'steelblue', 'orangered']
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# directory of data files from external sources
extDir = dataDir / 'external'
# parent directory of all processed data files
procDir = dataDir / 'processed'
# directory of processed data files specific to interactome
interactomeDir = procDir / interactome_name
# directory of GO similarity output data files
gosimDir = interactomeDir / 'gosim'
# figure directory
figDir = Path('../figures') / interactome_name / 'gosim'
# get root ontology label to label output files and figures
ont_label = ont_abv[ont_root]
# input data files
ontologyFile = extDir / 'go-basic.obo'
annotationFile = extDir / 'goa_human.gaf'
interactomeFile = interactomeDir / 'human_interactome.txt'
structuralInteractomeFile = interactomeDir / 'human_interface_annotated_interactome.txt'
# output data files
refPPIListFile = gosimDir / 'refPPIs.txt'
randPairListFile = gosimDir / 'randPairs.txt'
refPPIgosimParamFile = gosimDir / ('fastsemsim_parameters_refPPIs_%s_%s' %
(ont_label, sim_measure))
randPairgosimParamFile = gosimDir / (
'fastsemsim_parameters_randPairs_%s_%s' % (ont_label, sim_measure))
refPPIfastsemsimOutFile = gosimDir / ('fastsemsim_output_refPPIs_%s_%s' %
(ont_label, sim_measure))
randPairfastsemsimOutFile = gosimDir / (
'fastsemsim_output_randPairs_%s_%s' % (ont_label, sim_measure))
refPPIgosimFile = gosimDir / ('gosim_refPPIs_%s_%s.pkl' %
(ont_label, sim_measure))
randPairgosimFile = gosimDir / ('gosim_randPairs_%s_%s.pkl' %
(ont_label, sim_measure))
allPPIgosimFile = gosimDir / ('allPPI_gosim_%s_%s.pkl' %
(ont_label, sim_measure))
# create directories if not existing
if not gosimDir.exists():
os.makedirs(gosimDir)
if not figDir.exists():
os.makedirs(figDir)
#------------------------------------------------------------------------------------
# load reference and structural interactomes
#------------------------------------------------------------------------------------
interactome = pd.read_table(interactomeFile)
interactomeProteins = list(
set(interactome[["Protein_1", "Protein_2"]].values.flatten()))
print('\n' + 'reference interactome:')
print('%d PPIs' % len(interactome))
print('%d proteins' % len(interactomeProteins))
structuralInteractome = read_single_interface_annotated_interactome(
structuralInteractomeFile)
strucInteractomeProteins = list(
set(structuralInteractome[["Protein_1",
"Protein_2"]].values.flatten()))
print('\n' + 'interface-annotated interactome:')
print('%d PPIs' % len(structuralInteractome))
print('%d proteins' % len(strucInteractomeProteins))
#------------------------------------------------------------------------------------
# Calculate GO similarity for random interactions
#------------------------------------------------------------------------------------
randPairs = pd.DataFrame()
randPairs["Protein_1"], randPairs["Protein_2"] = zip(
*sample_random_pairs(interactomeProteins, numRandPairs))
randPairs = randPairs[randPairs["Protein_1"] != randPairs["Protein_2"]]
# produce protein GO similarity dictionary
print('\n' + 'producing GO similarity dictionary for random pairs')
randPairs[["Protein_1", "Protein_2"]].to_csv(randPairListFile,
index=False,
header=False,
sep='\t')
produce_fastsemsim_protein_gosim_dict(
randPairListFile,
randPairgosimFile,
sim_measure=sim_measure,
mix_method=mix_method,
ont_root=ont_root,
ont_ignore=ont_ignore,
ec_ignore=ec_ignore,
ontologyFile=ontologyFile,
annotationFile=annotationFile,
paramOutFile=randPairgosimParamFile,
fastsemsimOutFile=randPairfastsemsimOutFile)
with open(randPairgosimFile, 'rb') as f:
gosim = pickle.load(f)
sim = []
for p in map(tuple,
map(sorted, randPairs[["Protein_1", "Protein_2"]].values)):
sim.append(gosim[p] if p in gosim else np.nan)
randPairs["gosim"] = sim
randPairs = randPairs[np.isnan(randPairs["gosim"]) == False].reset_index(
drop=True)
#------------------------------------------------------------------------------------
# Calculate GO similarity for all interaction partners in the reference interactome
#------------------------------------------------------------------------------------
refPPIs = pd.DataFrame()
refPPIs["Protein_1"], refPPIs["Protein_2"] = zip(
*interactome[["Protein_1", "Protein_2"]].values)
# produce protein GO similarity dictionary
if not refPPIgosimFile.is_file():
print('\n' + 'producing GO similarity dictionary for reference PPIs')
refPPIs[["Protein_1", "Protein_2"]].to_csv(refPPIListFile,
index=False,
header=False,
sep='\t')
produce_fastsemsim_protein_gosim_dict(
refPPIListFile,
refPPIgosimFile,
sim_measure=sim_measure,
mix_method=mix_method,
ont_root=ont_root,
ont_ignore=ont_ignore,
ec_ignore=ec_ignore,
ontologyFile=ontologyFile,
annotationFile=annotationFile,
paramOutFile=refPPIgosimParamFile,
fastsemsimOutFile=refPPIfastsemsimOutFile)
with open(refPPIgosimFile, 'rb') as f:
gosim = pickle.load(f)
sim = []
for p in map(tuple, map(sorted, refPPIs[["Protein_1",
"Protein_2"]].values)):
sim.append(gosim[p] if p in gosim else np.nan)
refPPIs["gosim"] = sim
refPPIs = refPPIs[np.isnan(refPPIs["gosim"]) == False].reset_index(
drop=True)
#------------------------------------------------------------------------------------
# Calculate GO similarity for all interaction partners in the structural interactome
#------------------------------------------------------------------------------------
strucPPIs = pd.DataFrame()
strucPPIs["Protein_1"], strucPPIs["Protein_2"] = zip(
*structuralInteractome[["Protein_1", "Protein_2"]].values)
sim = []
for p in map(tuple,
map(sorted, strucPPIs[["Protein_1", "Protein_2"]].values)):
sim.append(gosim[p] if p in gosim else np.nan)
strucPPIs["gosim"] = sim
strucPPIs = strucPPIs[np.isnan(strucPPIs["gosim"]) == False].reset_index(
drop=True)
#------------------------------------------------------------------------------------
# Save GO similarity results to file
#------------------------------------------------------------------------------------
allgosim = {
k: sim
for k, sim in zip([
'Random interactions', 'Reference interactome',
'Structural interactome'
], [randPairs, refPPIs, strucPPIs])
}
with open(allPPIgosimFile, 'wb') as fout:
pickle.dump(allgosim, fout)
#------------------------------------------------------------------------------------
# Compare GO similarity of interaction partners between reference interactome,
# structural interactome and random interactions
#------------------------------------------------------------------------------------
# remove NaN values
randGOsim = randPairs["gosim"].tolist()
refGOsim = refPPIs["gosim"].tolist()
strucGOsim = strucPPIs["gosim"].tolist()
# print results
print('\n' + 'Mean %s similarity for interaction partners:' % ont_root)
print('Random pairs: %f (SE = %g, n = %d)' %
(np.mean(randGOsim), sderror(randGOsim), len(randGOsim)))
print('Reference interactome: %f (SE = %g, n = %d)' %
(np.mean(refGOsim), sderror(refGOsim), len(refGOsim)))
print('Structural interactome: %f (SE = %g, n = %d)' %
(np.mean(strucGOsim), sderror(strucGOsim), len(strucGOsim)))
print('\n' + 'Statistical significance')
print('reference interactome vs random interactions:')
bootstrap_test(refGOsim, randGOsim, iter=100000)
print('structural interactome vs reference interactome:')
bootstrap_test(strucGOsim, refGOsim, iter=100000)
# plot results
bar_plot([np.mean(randGOsim),
np.mean(refGOsim),
np.mean(strucGOsim)],
[sderror(randGOsim),
sderror(refGOsim),
sderror(strucGOsim)],
xlabels=[
'Random\ninteractions', 'Reference\ninteractome',
'Structural\ninteractome'
],
ylabel='%s similarity of\ninteraction partners' % ont_root,
colors=interactome_colors,
edgecolor='k',
barwidth=0.5,
fontsize=24,
show=showFigs,
figdir=figDir,
figname='interactome_gosim_%s_%s' % (ont_label, sim_measure))
def main():
# reference interactome name
# options: HI-II-14, IntAct
ref_interactome_name = 'HI-II-14'
# tissue expression databases
expr_db = ['Illumina', 'GTEx', 'HPA', 'Fantom5']
# structural interactome names
struc_name = {'HI-II-14': 'Y2H-SI', 'IntAct': 'IntAct-SI'}
# interactome names
interactome_names = [
'Random interactions', 'Reference interactome',
'Structural interactome'
]
# interactome colors
interactome_colors = ['limegreen', 'steelblue', 'orangered']
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# parent directory of processed data files
procDir = dataDir / 'processed'
# directory of processed data files specific to interactome
interactomeDir = procDir / ref_interactome_name
# figure directory
figDir = Path('../figures') / ref_interactome_name / 'coexpr'
# input data files
coexprFiles = [
interactomeDir / 'coexpr' / ('interactome_coexpr_%s.pkl' % db)
for db in expr_db
]
# create output directories if not existing
if not figDir.exists():
os.makedirs(figDir)
allcoexpr = {}
for db, coexprFile in zip(expr_db, coexprFiles):
with open(coexprFile, 'rb') as f:
allcoexpr[db] = pickle.load(f)
means, errors = [], []
for interactome_name in interactome_names:
coexpr = [allcoexpr[db][interactome_name]["coexpr"] for db in expr_db]
means.append([np.mean(c) for c in coexpr])
errors.append([sderror(c) for c in coexpr])
multi_bar_plot(means,
errors=errors,
xlabels=expr_db,
ylabel='Tissue co-expression of\ninteraction partners',
ylabels=[round(x, 1) for x in np.arange(0, 0.9, 0.2)],
colors=interactome_colors,
edgecolor='k',
ewidth=1.5,
barwidth=0.2,
bargap=0.03,
fontsize=18,
capsize=5,
msize=8,
leg=[
'Random pairs',
'%s reference interactome' % ref_interactome_name,
struc_name[ref_interactome_name]
],
show=showFigs,
figdir=figDir,
figname='interactome_coexpr')
def main():
# substitution matrix name
matrixName = 'PAM30'
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# parent directory of all processed data files
procDir = dataDir / 'processed'
# figure directory
figDir = Path('../figures') / 'combined'
# input data files
naturalMutationsFile = procDir / 'dbsnp_mutations4.txt'
diseaseMutationsFile = procDir / 'clinvar_mutations6.txt'
# output data files
subsMatrixFile = procDir / (matrixName + '.pkl')
# create output directories if not existing
if not procDir.exists():
os.makedirs(procDir)
if not figDir.exists():
os.makedirs(figDir)
#------------------------------------------------------------------------------------
# produce substitution matrix
#------------------------------------------------------------------------------------
if not subsMatrixFile.is_file():
produce_substitution_matrix (matrixName, subsMatrixFile)
with open(subsMatrixFile, 'rb') as f:
subsTable = pickle.load(f)
#------------------------------------------------------------------------------------
# load mutations
#------------------------------------------------------------------------------------
naturalMutations, diseaseMutations = remove_mutation_overlaps (naturalMutationsFile, diseaseMutationsFile)
#------------------------------------------------------------------------------------
# calculate mutation substitution scores
#------------------------------------------------------------------------------------
natMutScore = [subsTable[x] for x in zip(naturalMutations["wt_res"], naturalMutations["mut_res"])]
disMutScore = [subsTable[x] for x in zip(diseaseMutations["wt_res"], diseaseMutations["mut_res"])]
print()
print( 'Avg. score for natural mutations: %.1f (SE = %g, n = %d)' % (np.mean(natMutScore),
sderror(natMutScore),
len(natMutScore)) )
print( 'Avg. score for disease mutations: %.1f (SE = %g, n = %d)' % (np.mean(disMutScore),
sderror(disMutScore),
len(disMutScore)) )
bootstrap_test (natMutScore, disMutScore, 10000)
box_plot([natMutScore, disMutScore],
xlabels = ('Non-disease\nmutations','Disease\nmutations'),
ylabels = [-15, -10, -5, 0, 5],
ylabel = 'PAM30 substitution score',
ylim = [-16, 5],
colors = ['turquoise', 'magenta'],
fontsize = 26,
show = showFigs,
figdir = figDir,
figname = 'substitution_score')
def main():
# reference interactome name
# options: HI-II-14, IntAct
interactome_name = 'HI-II-14'
# tissue expression database name
# options: Illumina, GTEx, HPA, Fantom5
expr_db = 'Illumina'
# minimum number of tissue expression values required for protein pair tissue
# co-expression to be considered
coexprMinTissues = 5
# number of random interactions
numRandPairs = 10000
# interactome colors
interactome_colors = ['limegreen', 'steelblue', 'orangered']
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# directory of data files from external sources
extDir = dataDir / 'external'
# parent directory of all processed data files
procDir = dataDir / 'processed'
# directory of processed data files specific to interactome
interactomeDir = procDir / interactome_name
# directory of tissue coexpression output data files
coexprDir = interactomeDir / 'coexpr'
# figure directory
figDir = Path('../figures') / interactome_name / 'coexpr'
# input data files
illuminaExprFile = extDir / 'E-MTAB-513.tsv.txt'
gtexDir = extDir / 'GTEx_Analysis_v7_eQTL_expression_matrices'
hpaExprFile = extDir / 'normal_tissue.tsv'
fantomExprFile = extDir / 'hg38_fair+new_CAGE_peaks_phase1and2_tpm_ann.osc.txt'
fantomSampleTypeFile = extDir / 'fantom5_sample_type.xlsx'
uniprotIDmapFile = procDir / 'to_human_uniprotID_map.pkl'
uniqueGeneSwissProtIDFile = procDir / 'uniprot_unique_gene_reviewed_human_proteome.list'
interactomeFile = interactomeDir / 'human_interactome.txt'
structuralInteractomeFile = interactomeDir / 'human_interface_annotated_interactome.txt'
# output data files
proteinExprFile = procDir / ('protein_expr_%s.pkl' % expr_db)
coexprFile = coexprDir / ('interactome_coexpr_%s.pkl' % expr_db)
# create directories if not existing
if not coexprDir.exists():
os.makedirs(coexprDir)
if not figDir.exists():
os.makedirs(figDir)
#------------------------------------------------------------------------------------
# load reference and structural interactomes
#------------------------------------------------------------------------------------
interactome = pd.read_table(interactomeFile)
print('\n' + 'reference interactome:')
print('%d PPIs' % len(interactome))
print('%d proteins' %
len(set(interactome[["Protein_1", "Protein_2"]].values.flatten())))
structuralInteractome = read_single_interface_annotated_interactome(
structuralInteractomeFile)
print('\n' + 'interface-annotated interactome:')
print('%d PPIs' % len(structuralInteractome))
print('%d proteins' % len(
set(structuralInteractome[["Protein_1", "Protein_2"
]].values.flatten())))
#------------------------------------------------------------------------------------
# Produce tissue expression dictionary
#------------------------------------------------------------------------------------
# produce protein tissue expression profiles
if not proteinExprFile.is_file():
print('\n' + 'producing protein tissue expression dictionary')
if expr_db is 'Illumina':
produce_illumina_expr_dict(illuminaExprFile,
uniprotIDmapFile,
proteinExprFile,
headers=list(range(1, 18)))
elif expr_db is 'GTEx':
produce_gtex_expr_dict(gtexDir,
uniprotIDmapFile,
proteinExprFile,
uniprotIDlistFile=uniqueGeneSwissProtIDFile)
elif expr_db is 'HPA':
produce_hpa_expr_dict(hpaExprFile, uniprotIDmapFile,
proteinExprFile)
elif expr_db is 'Fantom5':
produce_fantom5_expr_dict(
fantomExprFile,
uniprotIDmapFile,
proteinExprFile,
sampleTypes='tissues',
sampleTypeFile=fantomSampleTypeFile,
uniprotIDlistFile=uniqueGeneSwissProtIDFile)
with open(proteinExprFile, 'rb') as f:
expr = pickle.load(f)
if expr_db is 'HPA':
exprMap = {'Not detected': 0, 'Low': 1, 'Medium': 2, 'High': 3}
for k, v in expr.items():
for i, e in enumerate(v):
v[i] = exprMap[e] if e in exprMap else np.nan
expr[k] = np.array(v)
#-----------------------------------------------------------------------------------------
# Calculate tissue co-expression for random interactions
#-----------------------------------------------------------------------------------------
proteins = list(
set(interactome[["Protein_1", "Protein_2"]].values.flatten()))
randPairs = pd.DataFrame()
randPairs["Protein_1"], randPairs["Protein_2"] = zip(
*sample_random_pairs(proteins, numRandPairs))
randPairs = randPairs[randPairs["Protein_1"] != randPairs["Protein_2"]]
randPairs["coexpr"] = randPairs.apply(lambda x: coexpr(
x["Protein_1"], x["Protein_2"], expr, minTissues=coexprMinTissues),
axis=1)
randPairs = randPairs[np.isnan(randPairs["coexpr"]) == False].reset_index(
drop=True)
#-----------------------------------------------------------------------------------------
# Calculate tissue co-expression for all interaction partners in the reference interactome
#-----------------------------------------------------------------------------------------
refPPIs = pd.DataFrame()
refPPIs["Protein_1"] = interactome["Protein_1"].tolist()
refPPIs["Protein_2"] = interactome["Protein_2"].tolist()
refPPIs["coexpr"] = refPPIs.apply(lambda x: coexpr(
x["Protein_1"], x["Protein_2"], expr, minTissues=coexprMinTissues),
axis=1)
refPPIs = refPPIs[np.isnan(refPPIs["coexpr"]) == False].reset_index(
drop=True)
#------------------------------------------------------------------------------------------
# Calculate tissue co-expression for all interaction partners in the structural interactome
#------------------------------------------------------------------------------------------
strucPPIs = pd.DataFrame()
strucPPIs["Protein_1"] = structuralInteractome["Protein_1"].tolist()
strucPPIs["Protein_2"] = structuralInteractome["Protein_2"].tolist()
strucPPIs["coexpr"] = strucPPIs.apply(lambda x: coexpr(
x["Protein_1"], x["Protein_2"], expr, minTissues=coexprMinTissues),
axis=1)
strucPPIs = strucPPIs[np.isnan(strucPPIs["coexpr"]) == False].reset_index(
drop=True)
#-------------------------------------------------------------------------------------
# Save tissue co-expression results to file
#-------------------------------------------------------------------------------------
allcoexpr = {
k: c
for k, c in zip([
'Random interactions', 'Reference interactome',
'Structural interactome'
], [randPairs, refPPIs, strucPPIs])
}
with open(coexprFile, 'wb') as fout:
pickle.dump(allcoexpr, fout)
#------------------------------------------------------------------------------------
# Compare tissue co-expression of interaction partners between reference interactome,
# structural interactome and random interactions
#------------------------------------------------------------------------------------
# remove NaN values
randCoexpr = randPairs["coexpr"].tolist()
refCoexpr = refPPIs["coexpr"].tolist()
strucCoexpr = strucPPIs["coexpr"].tolist()
# print results
print('\n' + 'Mean tissue co-expression for interaction partners:')
print('Random pairs: %f (SE = %g, n = %d)' %
(np.mean(randCoexpr), sderror(randCoexpr), len(randCoexpr)))
print('Reference interactome: %f (SE = %g, n = %d)' %
(np.mean(refCoexpr), sderror(refCoexpr), len(refCoexpr)))
print('Structural interactome: %f (SE = %g, n = %d)' %
(np.mean(strucCoexpr), sderror(strucCoexpr), len(strucCoexpr)))
print('\n' + 'Statistical significance')
print('reference interactome vs random interactions:')
bootstrap_test(refCoexpr, randCoexpr, iter=100000)
print('structural interactome vs reference interactome:')
bootstrap_test(strucCoexpr, refCoexpr, iter=100000)
# plot results
bar_plot([np.mean(randCoexpr),
np.mean(refCoexpr),
np.mean(strucCoexpr)],
[sderror(randCoexpr),
sderror(refCoexpr),
sderror(strucCoexpr)],
xlabels=[
'Random\ninteractions', 'Reference\ninteractome',
'Structural\ninteractome'
],
ylabel='Tissue co-expression of\ninteraction partners',
colors=interactome_colors,
edgecolor='k',
barwidth=0.5,
fontsize=24,
show=showFigs,
figdir=figDir,
figname='interactome_coexpression_%s' % expr_db)
def main():
# reference interactome name
# options: HI-II-14, IntAct
interactome_name = 'HI-II-14'
# structural interactome names
struc_name = {'HI-II-14': 'Y2H-SI', 'IntAct': 'IntAct-SI'}
# method of calculating mutation ∆∆G for which results will be used
# options: bindprofx, foldx
ddg_method = 'bindprofx'
# Minimum reduction in binding free energy DDG required for interaction perturbation
ddgCutoff = 0.5
# % confidence interval
CI = 95
# plot perturbed interactome and produce files for use by Cytoscape
plot_perturbations = False
# show figures
showFigs = False
# parent directory of all data files
dataDir = Path('../data')
# parent directory of all processed data files
procDir = dataDir / 'processed'
# directory of processed data files specific to interactome
interactomeDir = procDir / interactome_name
# directory of network perturbation output data files for use by Cytoscape
cytoscapeDir = interactomeDir / 'cytoscape'
# figure directory
figDir = Path('../figures') / interactome_name
# input data files
geometryPerturbsFile = interactomeDir / 'unique_mutation_perturbs_geometry.pkl'
natMutDDGinFile = interactomeDir / ('nondisease_mutations_%s_ddg.txt' %
ddg_method)
disMutDDGinFile = interactomeDir / ('disease_mutations_%s_ddg.txt' %
ddg_method)
interfaceAnnotatedInteractomeFile = interactomeDir / 'human_interface_annotated_interactome.txt'
# output data files
natMutDDGoutFile = interactomeDir / ('nondisMut_%s_∆∆G_used.txt' %
ddg_method)
disMutDDGoutFile = interactomeDir / ('disMut_%s_∆∆G_used.txt' % ddg_method)
physicsPerturbsFile = interactomeDir / (
'mutation_perturbs_physics_%s.pkl' % ddg_method)
naturalMutEdgeFile = cytoscapeDir / (
'nondiseaseMut_perturbed_edges_physics_%s' % ddg_method)
naturalMutNodeFile = cytoscapeDir / (
'nondiseaseMut_node_colors_physics_%s' % ddg_method)
diseaseMutEdgeFile = cytoscapeDir / (
'diseaseMut_perturbed_edges_physics_%s' % ddg_method)
diseaseMutNodeFile = cytoscapeDir / ('diseaseMut_node_colors_physics_%s' %
ddg_method)
# create output directories if not existing
if not interactomeDir.exists():
os.makedirs(interactomeDir)
if not cytoscapeDir.exists():
os.makedirs(cytoscapeDir)
if not figDir.exists():
os.makedirs(figDir)
#------------------------------------------------------------------------------------
# Fraction of mutation-targeted PPIs with ∆∆G exceeding a specified cutoff
#------------------------------------------------------------------------------------
# read change in binding free energy for interfacial mutations mapped on PDB chains
naturalMutationsDDG = read_protein_mutation_ddg(natMutDDGinFile, 'binding')
diseaseMutationsDDG = read_protein_mutation_ddg(disMutDDGinFile, 'binding')
naturalMutations = pd.DataFrame(columns=[
"protein", "partner", "protein_pos", "mut_res", "pdb_id", "chain_id",
"chain_partner", "chain_mut", "ddg"
])
for i, item in enumerate(naturalMutationsDDG.items()):
naturalMutations.loc[i] = item[0] + item[1]
diseaseMutations = pd.DataFrame(columns=[
"protein", "partner", "protein_pos", "mut_res", "pdb_id", "chain_id",
"chain_partner", "chain_mut", "ddg"
])
for i, item in enumerate(diseaseMutationsDDG.items()):
diseaseMutations.loc[i] = item[0] + item[1]
naturalMutations.to_csv(natMutDDGoutFile, index=False, sep='\t')
diseaseMutations.to_csv(disMutDDGoutFile, index=False, sep='\t')
natMutDDGs = naturalMutations["ddg"].values
disMutDDGs = diseaseMutations["ddg"].values
numNatural_ddg_considered = len(natMutDDGs)
numDisease_ddg_considered = len(disMutDDGs)
print('\n' +
'Avg. change in binding energy (∆∆G) for mutation-targeted PPIs:')
print(
'Non-disease: %.1f (SE = %g, n = %d)' %
(np.mean(natMutDDGs), sderror(natMutDDGs), numNatural_ddg_considered))
print(
'Disease: %.1f (SE = %g, n = %d)' %
(np.mean(disMutDDGs), sderror(disMutDDGs), numDisease_ddg_considered))
# Statistical significance of difference in means
t_test(natMutDDGs, disMutDDGs)
multi_histogram_plot([disMutDDGs, natMutDDGs], ['red', 'green'],
xlabel='Change in binding free energy (∆∆G)',
ylabel='Number of mutations',
leg=[
'Disease interfacial mutations',
'Non-disease interfacial mutations'
],
bins=25,
alpha=0.7,
fontsize=24,
show=showFigs,
figdir=figDir,
figname='mut_ddg_histogram_%s' % ddg_method)
numNatural_ddg = sum(natMutDDGs > ddgCutoff)
numDisease_ddg = sum(disMutDDGs > ddgCutoff)
fracNatural_ddg = numNatural_ddg / numNatural_ddg_considered
fracDisease_ddg = numDisease_ddg / numDisease_ddg_considered
fracNatural_ddg_error = sderror_on_fraction(numNatural_ddg,
numNatural_ddg_considered)
fracDisease_ddg_error = sderror_on_fraction(numDisease_ddg,
numDisease_ddg_considered)
print('\n' +
'Fraction of mutation-targeted PPIs with ∆∆G > %.1f kcal/mol:' %
ddgCutoff)
print('Non-disease: %.3f (SE = %g, ntot = %d)' %
(fracNatural_ddg, fracNatural_ddg_error, numNatural_ddg_considered))
print('Disease: %.3f (SE = %g, ntot = %d)' %
(fracDisease_ddg, fracDisease_ddg_error, numDisease_ddg_considered))
# Statistical significance of difference in fractions
fisher_test([numNatural_ddg, numNatural_ddg_considered - numNatural_ddg],
[numDisease_ddg, numDisease_ddg_considered - numDisease_ddg])
bar_plot([fracNatural_ddg, fracDisease_ddg],
error=[fracNatural_ddg_error, fracDisease_ddg_error],
xlabels=('%s PPIs\nwith non-disease\nmutations\nat interface' %
struc_name[interactome_name],
'%s PPIs\nwith disease\nmutations\nat interface' %
struc_name[interactome_name]),
ylabel=('Fraction with ∆∆G > %.1f kcal / mol' % ddgCutoff),
ylabels=[0, 0.2, 0.4, 0.6, 0.8],
ylim=[0, 0.8],
colors=['turquoise', 'magenta'],
edgecolor='black',
ewidth=2.5,
barwidth=0.6,
fontsize=24,
capsize=10,
msize=26,
show=showFigs,
figdir=figDir,
figname='mut_ddg_frac_>%.1f_%s' % (ddgCutoff, ddg_method))
#------------------------------------------------------------------------------------
# predict PPI perturbations
#------------------------------------------------------------------------------------
if geometryPerturbsFile.is_file():
print('\n' + 'Loading geometry-based PPI perturbation predictions')
with open(geometryPerturbsFile, 'rb') as f:
naturalPerturbs, diseasePerturbs = pickle.load(f)
else:
print('\n' +
'Geometry-based PPI perturbation prediction file not found')
return
print(
'\n' +
'Performing physics-based edgotype prediction for non-disease mutations'
)
naturalPerturbs[
"perturbations"], knownDDG, unknownDDG = energy_based_perturbation(
naturalPerturbs, naturalMutationsDDG, ddgCutoff)
print('\n' +
'Performing physics-based edgotype prediction for disease mutations')
diseasePerturbs[
"perturbations"], knownDDG, unknownDDG = energy_based_perturbation(
diseasePerturbs, diseaseMutationsDDG, ddgCutoff)
with open(physicsPerturbsFile, 'wb') as fOut:
pickle.dump([naturalPerturbs, diseasePerturbs], fOut)
#------------------------------------------------------------------------------------
# plot network perturbations
#------------------------------------------------------------------------------------
if plot_perturbations:
structuralInteractome = read_interface_annotated_interactome(
interfaceAnnotatedInteractomeFile)
print('\n' + 'Creating network perturbed by non-disease mutations')
nodes, edges, nodeColors, edgeColors = create_perturbed_network(
structuralInteractome, naturalPerturbs, naturalMutEdgeFile,
naturalMutNodeFile)
network_plot(
edges,
nodes=nodes,
nodeSizes=[20] * len(nodes),
edgeWidth=1,
nodeColors=nodeColors,
edgeColors=edgeColors,
show=showFigs,
figdir=figDir,
figname='nondisease_mut_perturbed_interactome_physics_%s' %
ddg_method)
print('\n' + 'Creating network perturbed by disease mutations')
nodes, edges, nodeColors, edgeColors = create_perturbed_network(
structuralInteractome, diseasePerturbs, diseaseMutEdgeFile,
diseaseMutNodeFile)
network_plot(edges,
nodes=nodes,
nodeSizes=[20] * len(nodes),
edgeWidth=1,
nodeColors=nodeColors,
edgeColors=edgeColors,
show=showFigs,
figdir=figDir,
figname='disease_mut_perturbed_interactome_physics_%s' %
ddg_method)