def processPathwayBKF(self, exhaustiveOr=False):
assert len(self.pathways) > 0, "Have not processed pathways yet"
for pathway in tqdm.tqdm(self.pathways,
desc='Processing pathway BKFs'):
bkf = BKB(name=pathway.pathwayID)
pathwayActiveComp_idx = bkf.addComponent('{}_active='.format(
pathway.pathwayID))
pathwayActiveTrue_idx = bkf.addComponentState(
pathwayActiveComp_idx, 'True')
if exhaustiveOr:
print("not implemented")
else:
geneSelectorComp_idx = bkf.addComponent("Gene_combo=")
for gene in pathway.genes: #gene[0] = geneName, gene[1] = low value gene[2] = high value
geneCombo_idx = bkf.addComponentState(
geneSelectorComp_idx, gene[0])
statConditionComp_idx = bkf.addComponent(
'mu-STD<={}<=mu+STD='.format(gene[0]))
statConditionTrue_idx = bkf.addComponentState(
statConditionComp_idx, 'True')
bkf.addSNode(
BKB_S_node(statConditionComp_idx,
statConditionTrue_idx, 1.0))
bkf.addSNode(
BKB_S_node(
geneSelectorComp_idx, geneCombo_idx, 1.0,
[(statConditionComp_idx, statConditionTrue_idx)]))
bkf.addSNode(
BKB_S_node(pathwayActiveComp_idx,
pathwayActiveTrue_idx, 1.0,
[(geneSelectorComp_idx, geneCombo_idx)]))
self.bkfs.append(bkf)
assert len(self.pathways) > 0, "Have not processed pathways yet"
query = self.reasoner.analyze_query(
copy.deepcopy(query), preprocessed_bkb=self.processed_bkb)
summary = query.result.summary()
#query.result.summary()
for update, prob in query.result.updates.items():
comp_idx, state_idx = update
comp_name = query.bkb.getComponentName(comp_idx)
state_name = query.bkb.getComponentINodeName(
comp_idx, state_idx)
probs.append((comp_name, state_name, prob))
return (probs[0], probs[1]), summary
if __name__ == '__main__':
fused_bkb = BKB()
fused_bkb.load('/home/public/data/ncats/663Pats6Holdouts/fusion.bkb')
patient_data_file = '/home/public/data/ncats/663Pats6Holdouts/patient_data.pk'
withheld_patients_file = '/home/public/data/ncats/663Pats6Holdouts/withheldPatients.csv'
#fused_bkb.load('/home/public/data/ncats/90PERCENTValidation50Patients10Validation/set1/fusion.bkb')
#patient_data_file = '/home/public/data/ncats/90PERCENTValidation50Patients10Validation/set1/patient_data.pk'
#withheld_patients_file = '/home/public/data/ncats/90PERCENTValidation50Patients10Validation/set1/withheldPatients.csv'
compNames = fused_bkb.getAllComponentNames()
f = open(withheld_patients_file, 'r')
withheldPatientHashes = f.read().split(',')
patientDict = pickle.load(open(patient_data_file, 'rb'))
patientsDynamicEvidence = []
priors[gene] = float(count / len(PATIENTS))
#--Pairs Calculate Correlations
counts = dict()
for gene1 in GENES:
for gene2 in GENES:
count = 0
for patient in PATIENTS:
if patient_data[patient]['Genes'][gene1] and patient_data[patient][
'Genes'][gene2]:
count += 1
counts[(gene1, gene2)] = float(count / len(PATIENTS)) / priors[gene2]
#-- Make BKB Patient frags
for patient in PATIENTS:
bkf = BKB()
for gene in patient_data[patient]['Genes']:
#-- Setup Gene i component.
geneComp_idx = bkf.addComponent(gene)
geneStateTrue_idx = bkf.addComponentState(geneComp_idx, 'True')
s_node_prior = BKB_S_node(init_component_index=geneComp_idx,
init_state_index=geneStateTrue_idx,
init_probability=1)
bkf.addSNode(s_node_prior)
patient_bkfs.append(bkf)
#-- Define Sample Pathway
PATHWAYS = ['Pathway{}'.format(i) for i in range(1)]
AGE_RANGE_YRS = (20, 90)
SURIVAL_YRS = (0, 10)
patient_data = {
patient: {
'Gender': random.choice(GENDER_OPTIONS),
'Age': random.randrange(AGE_RANGE_YRS[0], AGE_RANGE_YRS[1]),
'Survival': random.randrange(SURIVAL_YRS[0], SURIVAL_YRS[1])
}
for patient in PATIENTS
}
bkfs = list()
#-- Make BKB frags
for patient in PATIENTS:
bkf = BKB()
for gene in GENES:
if random.choice([True, False]):
#-- Setup Gene i component.
comp_idx = bkf.addComponent('mut_{}'.format(gene))
stateTrue_idx = bkf.addComponentState(comp_idx, 'True')
bkf.addSNode(
BKB_S_node(init_component_index=comp_idx,
init_state_index=stateTrue_idx,
init_probability=1))
variant_comp_idx = bkf.addComponent('mut-var_{}'.format(gene))
random_variant = random.choice(GENE_VARIANTS)
variant_state_idx = bkf.addComponentState(variant_comp_idx,
random_variant)
import os
import sys
from pybkb import bayesianKnowledgeBase as BKB
#-- Change to your local data folder
NCATS_DIR = '/home/cyakaboski/src/python/pojects/bkb-pathway-provider'
sys.path.append(os.path.join(NCATS_DIR, 'core'))
from query import Query
from reasoner import Reasoner
if __name__ == '__main__':
bkb = BKB()
bkb.load('/home/public/data/ncats/fusedPatientBKBSmall/fusion.bkb')
reasoner = Reasoner(bkb)
query0 = Query(evidence={
'_sourceFusion_14_02_2020_16:35:04__mut_CACNA1H=':
'[0]_-7750956075882572850'
},
targets=['mut_CACNA1H=', 'mu-STD>=CACNA1H<=mu+STD='],
type='updating')
result = reasoner.analyze_query(query0)
def read_expression_snodes(self, expression_levels_dict, tcga_expression_file, react_gene_file, pathways_file):
#-- Process all gene_reaction s_nodes and assume there is a header with gene names.
with open(react_gene_file, 'r') as csv_file:
reader = csv.reader(csv_file)
rows = [row for row in reader]
#-- Process Gene Names from header:
print(rows[0])
genes = rows[0][1:]
genes = [genes[i][:-4] for i in range(0,len(genes),2)]
print(genes)
for row in rows[1:]:
bkf = BKB(name=row[0] + ' BKF')
reaction_head_name = row[0]
react_component = BKB_component(reaction_head_name)
i_node_true = BKB_I_node('True', react_component)
i_node_false = BKB_I_node('False', react_component)
react_component.addINode(i_node_true)
react_component.addINode(i_node_false)
bkf.addComponent(react_component)
#-- Extract and Make I-nodes
for gene_j, row_i in enumerate(range(1,len(row), 2)):
gene_name = genes[gene_j]
gene_low = row[row_i]
gene_high = row[row_i + 1]
if gene_low != '':
print(gene_low, gene_high)
num_levels = expression_levels_dict[gene_name]
levels = [rag for rag in np.linspace(int(gene_low), int(gene_high), num_levels)]
component_states = ['{} | {}'.format(levels[i], levels[i+1]) for i in range(len(levels)-1)]
component = BKB_component(gene_name + ' ExpressionLv')
for state in component_states:
i_node = BKB_I_node(state, component)
component.addINode(i_node)
bkf.addComponent(component)
for idx in range(bkf.findComponent(component.name).getNumberStates()):
i_node = component.getState(idx)
bkf.addSNode(BKB_S_node(component, i_node, random.random()))
bkf.addSNode(BKB_S_node(react_component, i_node_true, random.random(), [(component, i_node)]))
bkf.addSNode(BKB_S_node(react_component, i_node_false, random.random(), [(component, i_node)]))
self.bkfs.append(bkf)
print('Done')
#-- Process Pathway BKFs
with open(pathways_file, 'r') as csv_file:
reader = csv.reader(csv_file)
bkf = BKB()
for row in reader:
tail = row[0]
head = row[1]
states = ['True', 'False']
if bkf.findComponent(tail) == -1:
component_tail = BKB_component(tail)
for state in states:
component_tail.addINode(BKB_I_node(state, component_tail))
bkf.addComponent(component_tail)
else:
component_tail = bkf.findComponent(tail)
if bkf.findComponent(head) == -1:
component_head = BKB_component(head)
for state in states:
component_head.addINode(BKB_I_node(state, component_head))
bkf.addComponent(component_head)
else:
component_head = bkf.findComponent(head)
for state1 in states:
for state2 in states:
bkf.addSNode(BKB_S_node(component_head,
component_head.findState(state1),
random.random(),
[(component_tail, component_tail.findState(state2))])
)
self.bkfs.append(bkf)
print('Complete')
def processPathwayBKF(self, exhaustiveOr=False):
assert len(self.pathways) > 0, "Have not processed pathways yet"
for pathway in tqdm.tqdm(self.pathways,
desc='Processing pathway BKFs'):
bkf = BKB(name=pathway.ID)
if "_hier" in pathway.ID:
#head
pathwayParentComp_idx = bkf.addComponent('{}_active='.format(
pathway.head))
pathwayParentTrue_idx = bkf.addComponentState(
pathwayParentComp_idx, 'True')
#tail - should only be 1 for _hier bkfs
pathwayChildComp_idx = bkf.addComponent('{}_active='.format(
pathway.tails[0]))
pathwayChildTrue_idx = bkf.addComponentState(
pathwayChildComp_idx, 'True')
#S-node
bkf.addSNode(
BKB_S_node(pathwayChildComp_idx, pathwayChildTrue_idx,
1.0))
bkf.addSNode(
BKB_S_node(pathwayParentComp_idx, pathwayParentTrue_idx,
1.0,
[(pathwayChildComp_idx, pathwayChildTrue_idx)]))
else:
'''
#head
pathwayHierComp_idx = bkf.addComponent('{}_active='.format(pathway.head))
pathwayHierTrue_idx = bkf.addComponentState(pathwayHierComp_idx, 'True')
#tails
for tail in pathway.tails:
statConditionComp_idx = bkf.addComponent('mu-STD<={}<=mu+STD'.format(tail))
statConditionTrue_idx = bkf.addComponentState(statConditionComp_idx, 'True')
bkf.addSNode(BKB_S_node(statConditionComp_idx, statConditionTrue_idx, 1.0))
bkf.addSNode(BKB_S_node(pathwayHierComp_idx, pathwayHierTrue_idx, 1.0, [(statConditionComp_idx, statConditionTrue_idx)]))
self.bkfs.append(bkf)
'''
#=======
if exhaustiveOr:
print("not implemented")
else:
# pathway
pathwayReactionComp_idx = bkf.addComponent(pathway.head +
"_active=")
pathwayReactionTrue_idx = bkf.addComponentState(
pathwayReactionComp_idx, 'True')
# gene slector
geneSelectorComp_idx = bkf.addComponent("Gene_combo=")
#tails
for tail in pathway.tails:
geneCombo_idx = bkf.addComponentState(
geneSelectorComp_idx, tail)
statConditionComp = BKB_component("mu-STD>=" + tail +
"<=mu+STD=")
statConditionTrue = BKB_I_node('True',
statConditionComp)
statConditionComp_idx = bkf.addComponent("mu-STD>=" +
tail +
"<=mu+STD=")
statConditionTrue_idx = bkf.addComponentState(
statConditionComp_idx, 'True')
bkf.addSNode(
BKB_S_node(statConditionComp_idx,
statConditionTrue_idx, 1.0))
bkf.addSNode(
BKB_S_node(geneSelectorComp_idx, geneCombo_idx,
1.0, [(statConditionComp_idx,
statConditionTrue_idx)]))
bkf.addSNode(
BKB_S_node(
pathwayReactionComp_idx,
pathwayReactionTrue_idx, 1.0,
[(geneSelectorComp_idx, geneCombo_idx)]))
self.bkfs.append(bkf)
def processPatientBKF(self):
assert len(
self.patients) > 0, "Have not processed Patient and gene data yet."
#print("Forming Patient BKFs...")
allGenes = list()
for pat in tqdm.tqdm(self.patients, desc='Forming patient BKFs'):
bkf = BKB(name=pat.patientID)
variantCompIndices = list()
variantCompNames = list()
variantCompStateIndices = list()
for idx, gene in enumerate(pat.mutatedGenes):
allGenes.append(gene)
# gene
mutGeneComp_idx = bkf.addComponent('mut_{}='.format(gene))
iNodeGeneMut_idx = bkf.addComponentState(
mutGeneComp_idx, 'True')
# form SNode o---->[mut_<genename>=True]
bkf.addSNode(BKB_S_node(mutGeneComp_idx, iNodeGeneMut_idx,
1.0))
if 'var_' + pat.variants[idx] + '=' in variantCompNames:
cIdx = variantCompNames.index('var_{}='.format(
pat.variants[idx]))
#mut_var combo
mutVarComp_idx = bkf.addComponent('mut-var_{}='.format(
pat.mutatedGeneVariants[idx]))
iNodeMutVar_idx = bkf.addComponentState(
mutVarComp_idx, 'True')
#mut_var s-node
bkf.addSNode(
BKB_S_node(mutVarComp_idx, iNodeMutVar_idx, 1.0,
[(mutGeneComp_idx, iNodeGeneMut_idx),
(variantCompIndices[cIdx],
variantCompStateIndices[cIdx])]))
else:
#var
variantComp_idx = bkf.addComponent('var_{}='.format(
pat.variants[idx]))
iNodeVariant_idx = bkf.addComponentState(
variantComp_idx, 'True')
#var s-node
bkf.addSNode(
BKB_S_node(variantComp_idx, iNodeVariant_idx, 1.0))
variantCompIndices.append(variantComp_idx)
variantCompStateIndices.append(iNodeGeneMut_idx)
variantCompNames.append(
bkf.getComponentName(variantComp_idx))
#mut_var combo
mutVarComp_idx = bkf.addComponent('mut-var_{}='.format(
pat.mutatedGeneVariants[idx]))
iNodeMutVar_idx = bkf.addComponentState(
mutVarComp_idx, 'True')
#mut_var s-node
bkf.addSNode(
BKB_S_node(mutVarComp_idx, iNodeMutVar_idx, 1.0,
[(mutGeneComp_idx, iNodeGeneMut_idx),
(variantComp_idx, iNodeVariant_idx)]))
self.bkfs.append(bkf)