def comet(mutations, n, t, ks, numIters, stepLen, initialSoln, amp, subt, nt,
hybridPvalThreshold, pvalThresh, verbose):
# Convert mutation data to C-ready format
if subt: mutations = mutations + (subt, )
cMutations = C.convert_mutations_to_C_format(*mutations)
iPatientToGenes, iGeneToCases, geneToNumCases, geneToIndex, indexToGene = cMutations
initialSolnIndex = [geneToIndex[g] for g in initialSoln]
solns = C.comet(t, mutations[0], mutations[1], iPatientToGenes,
geneToNumCases, ks, numIters,
stepLen, amp, nt, hybridPvalThreshold, initialSolnIndex,
len(subt), pvalThresh, verbose)
# Collate the results and sort them descending by sampling frequency
solnsWithWeights = C.convert_solns(indexToGene, solns)
def collection_key(collection):
return " ".join(sorted([",".join(sorted(M)) for M in collection]))
results = dict()
# store last soln of sampling for more iterations
lastSoln = list()
for gset in solnsWithWeights[-1][0]:
for g in gset:
lastSoln.append(g)
for collection, Ws, Cs in solnsWithWeights:
key = collection_key(collection)
if key in results: results[key]["freq"] += 1
else:
sets = []
for i in range(len(collection)):
M = collection[i]
W = Ws[i]
F = Cs[i]
# extract the probability from the weight,
# which can also include the accelerator
P = pow(exp(-W), 1. / amp)
sets.append(dict(genes=M, W=W, num_tbls=F, prob=P))
totalWeight = sum([S["W"] for S in sets])
targetWeight = exp(totalWeight) if totalWeight < 700 else 1e1000
results[key] = dict(freq=1,
sets=sets,
total_weight=totalWeight,
target_weight=targetWeight)
return results, lastSoln
def run(args):
# Parse the arguments into shorter variable hadnles
mutationMatrix = args.mutation_matrix
geneFile = args.gene_file
patientFile = args.patient_file
minFreq = args.min_freq
k = args.gene_set_size
pvalThresh = 1.1
wf = args.weight_func
# Load the mutation data
mutations = C.load_mutation_data(mutationMatrix, patientFile, geneFile,
minFreq)
m, n = mutations[0], mutations[1]
if args.verbose:
print('- Mutation data: %s genes x %s patients' % (m, n))
# Set up the CoMEt run and then run exhaustively
cMutations = C.convert_mutations_to_C_format(*mutations)
iPatientToGenes, iGeneToCases, geneToNumCases, geneToIndex, indexToGene = cMutations
genes = sorted(list(geneToIndex.keys()), key=lambda g: geneToIndex[g])
C.precompute_factorials(max(m, n))
C.set_weight(C.weightFunctionChars[wf])
results = C.exhaustive(k, m, n, iPatientToGenes, geneToNumCases,
pvalThresh)
C.free_factorials()
# Parse the output
solns, weights, tables, probs = results
res = list(zip(solns, weights, tables, probs))
res.sort(key=lambda arr: arr[1], reverse=True) # sort by weight decreasing
solns = [sorted([genes[g] for g in geneset]) for geneset, w, t, p in res]
weights = [w for g, w, t, p in res]
tables = [t for g, w, t, p in res]
probs = [p for g, w, t, p in res]
# Output only sets, probs, and freqs as TSV
with open("%s-k%s-%s-exhaustive.tsv" % (args.output_prefix, k, wf),
"w") as outfile:
output = [
"\t".join([", ".join(s), str(p), str(w)])
for s, p, w in zip(solns, probs, weights)
]
output.insert(0, "#Gene set\tP-value\tFreq\tWeight")
outfile.write("\n".join(output))
return list(zip(solns, probs, weights))
def comet(mutations, n, t, ks, numIters, stepLen, initialSoln,
amp, subt, nt, hybridPvalThreshold, pvalThresh, verbose):
# Convert mutation data to C-ready format
if subt: mutations = mutations + (subt, )
cMutations = C.convert_mutations_to_C_format(*mutations)
iPatientToGenes, iGeneToCases, geneToNumCases, geneToIndex, indexToGene = cMutations
initialSolnIndex = [geneToIndex[g] for g in initialSoln]
solns = C.comet(t, mutations[0], mutations[1], iPatientToGenes, geneToNumCases,
ks, numIters, stepLen, amp, nt, hybridPvalThreshold,
initialSolnIndex, len(subt), pvalThresh, verbose)
# Collate the results and sort them descending by sampling frequency
solnsWithWeights = convert_solns( indexToGene, solns )
def collection_key(collection):
return " ".join(sorted([",".join(sorted(M)) for M in collection]))
results = dict()
# store last soln of sampling for more iterations
lastSoln = list()
for gset in solnsWithWeights[-1][0]:
for g in gset:
lastSoln.append(g)
for collection, Ws, Cs in solnsWithWeights:
key = collection_key(collection)
if key in results: results[key]["freq"] += 1
else:
sets = []
for i in range(len(collection)):
M = collection[i]
W = Ws[i]
F = Cs[i]
# extract the probability from the weight,
# which can also include the accelerator
P = pow(exp(-W), 1./amp)
sets.append( dict(genes=M, W=W, num_tbls=F, prob=P) )
totalWeight = sum([ S["W"] for S in sets ])
targetWeight = exp( totalWeight ) if totalWeight < 700 else 1e1000
results[key] = dict(freq=1, sets=sets, total_weight=totalWeight,
target_weight=targetWeight)
return results, lastSoln
def run( args ):
# Parse the arguments into shorter variable hadnles
mutationMatrix = args.mutation_matrix
geneFile = args.gene_file
patientFile = args.patient_file
minFreq = args.min_freq
k = args.gene_set_size
pvalThresh = 1.1
wf = args.weight_func
# Load the mutation data
mutations = C.load_mutation_data(mutationMatrix, patientFile, geneFile, minFreq)
m, n = mutations[0], mutations[1]
if args.verbose:
print '- Mutation data: %s genes x %s patients' % (m, n)
# Set up the CoMEt run and then run exhaustively
cMutations = C.convert_mutations_to_C_format(*mutations)
iPatientToGenes, iGeneToCases, geneToNumCases, geneToIndex, indexToGene = cMutations
genes = sorted(geneToIndex.keys(), key=lambda g: geneToIndex[g])
C.precompute_factorials(max(m, n))
C.set_weight(C.weightFunctionChars[wf])
results = C.exhaustive(k, m, n, iPatientToGenes, geneToNumCases, pvalThresh)
C.free_factorials()
# Parse the output
solns, weights, tables, probs = results
res = zip(solns, weights, tables, probs)
res.sort(key=lambda arr: arr[1], reverse=True) # sort by weight decreasing
solns = [ sorted([genes[g] for g in geneset]) for geneset, w, t, p in res]
weights = [ w for g, w, t, p in res]
tables = [ t for g, w, t, p in res]
probs = [ p for g, w, t, p in res]
# Output only sets, probs, and freqs as TSV
with open("%s-k%s-%s-exhaustive.tsv" % (args.output_prefix, k, wf), "w") as outfile:
output = [ "\t".join([ ", ".join(s), str(p), str(w)])
for s, p, w in zip(solns, probs, weights)]
output.insert(0, "#Gene set\tP-value\tFreq\tWeight")
outfile.write( "\n".join(output) )
return zip(solns, probs, weights)
def load_precomputed_scores(infile, mutations, subt):
if subt: mutations = mutations + (subt,)
cMutations = C.convert_mutations_to_C_format(*mutations)
iPatientToGenes, iGeneToCases, geneToNumCases, geneToIndex, indexToGene = cMutations
baseI = 3 # sampling freq., total weight, target weight
setI = 3 # gene set, score, weight function
matchObj = re.match( r'.+\.k(\d+)\..+?', infile)
loadingT = len(matchObj.group(1)) # determine t:the number of gene sets.
for l in open(infile):
if not l.startswith("#"):
v = l.rstrip().split("\t")
j = 0
for i in range(loadingT):
gSet = [geneToIndex[g] for g in v[baseI + j].split(", ")]
C.load_precomputed_scores(float(v[baseI + j + 1]), len(v[baseI + j].split(", ")), int(v[baseI + j + 2]), gSet)
j += setI
