# These custom-written modules should have been included with the package
# distribution.
import graphFunctions as gf
import metadataFunctions as mf
import sbmlFunctions as sf
import seedFunctions as ef
# Define local folder structure for data input and processing.
processedDataDir = 'ProcessedModelFiles'
summaryStatsDir = 'DataSummaries'
externalDataDir = 'ExternalData'
#%% Program execution
# Import the list of models
dirList = mf.getDirList('../'+processedDataDir)
numSubDir = len(dirList)
# Convert SBML model files to adjacency lists
modelStatArray = sf.dirListToAdjacencyList(dirList, externalDataDir, processedDataDir, summaryStatsDir)
# Compute statistics on the size of metabolic network graphs
graphStatArray, diGraphStatArray = gf.computeGraphStats(dirList, processedDataDir, summaryStatsDir)
gf.plotGraphStats(graphStatArray)
# Reduce network graphs to their largest component
reducedGraphStatArray = gf.reduceToLargeComponent(dirList, processedDataDir, summaryStatsDir)
# Compute seed sets
seedSetList = gf.computeSeedSets(dirList, externalDataDir, processedDataDir)
gf.plotSeedStats(seedSetList, reducedGraphStatArray, modelStatArray)
def processSBMLforRE(rawModelDir, processedDataDir, summaryStatsDir,
externalDataDir):
# # Check that folders exist and create them if necessary
if not os.path.exists('../' + processedDataDir):
os.makedirs('../' + processedDataDir)
if not os.path.exists('../' + summaryStatsDir):
os.makedirs('../' + summaryStatsDir)
# Import the list of models
dirList = mf.getDirList('../' + rawModelDir)
numSubDir = len(dirList)
# Import the list of metabolies to revise
metabFormDict = {}
metabChargeDict = {}
with open('../' + externalDataDir + '/newFormulaDict.txt') as inFile:
for line in inFile:
(key, value) = line.split('\t')
metabFormDict[key] = value
with open('../' + externalDataDir + '/newChargeDict.txt') as inFile:
for line in inFile:
(key, value) = line.split('\t')
metabChargeDict[key] = value
# Create an array to store results
# Columns: genes, metabs, rxns, balanced (binary)
modelSizeDF = pd.DataFrame(
index=dirList,
columns=['Genes', 'Metabolites', 'Reactions', 'Balanced'])
# Intialize a counter
count = 1
for curDir in dirList:
# Print the subdirectory name
print 'Processing model ' + curDir + ', ' + str(count) + ' of ' + str(
numSubDir)
# Import metabolite charges
cpdData = pd.read_csv('../' + rawModelDir + '/' + curDir + '/' +
curDir + 'Compounds.tsv',
delimiter='\t',
index_col=0)
################################################################################
# Before reading in the SBML file, update gene loci so they read in properly
# In the old models ...
# KBase generates gene loci of the form kb|g.######.CDS.###
# Transform them to be of the form curDir_CDS_###
# In the new models ...
# KBase generates gene loci of the form curDir.genome.CDS.###
# Transform them to be of the form curDir_CDS_###
for myLine in fileinput.FileInput('../' + rawModelDir + '/' + curDir +
'/' + curDir + '.xml',
inplace=True):
# print re.sub('kb\|g\.\d+\.CDS\.(\d+)', curDir+'_CDS_\g<1>', myLine).strip()
print re.sub(curDir + '\.genome\.CDS\.(\d+)',
curDir + '_CDS_\g<1>', myLine).strip()
fileinput.close()
################################################################################
# Read in model from SBML
model = cobra.io.read_sbml_model('../' + rawModelDir + '/' + curDir +
'/' + curDir + '.xml')
################################################################################
# Remove undesired reactions, including:
badRxnList = []
# print 'Removing bad reactions'
for curRxn in model.reactions:
# Exchange reactions
if re.match('EX_', curRxn.id):
badRxnList.append(curRxn)
# Reactions w/o GPRs
elif len(curRxn.gene_reaction_rule) == 0:
badRxnList.append(curRxn)
# Protein, DNA, and RNA synthesis
elif curRxn.id == 'rxn13782_c0' or curRxn.id == 'rxn13783_c0' or curRxn.id == 'rxn13784_c0':
badRxnList.append(curRxn)
# Spontaneous reactions, whose GPR is fully 'unknown'
elif curRxn.gene_reaction_rule == 'Unknown':
badRxnList.append(curRxn)
# Transport reactions, based on keywords
elif re.search('transport', curRxn.name) or re.search(
'permease', curRxn.name) or re.search(
'symport', curRxn.name) or re.search(
'diffusion', curRxn.name) or re.search(
'excretion', curRxn.name) or re.search(
'export', curRxn.name) or re.search(
'secretion', curRxn.name) or re.search(
'uptake',
curRxn.name) or re.search(
'antiport', curRxn.name):
badRxnList.append(curRxn)
# Transport reactions which don't get picked up based on keywords
elif curRxn.id == 'rxn05226_c0' or curRxn.id == 'rxn05292_c0' or curRxn.id == 'rxn05305_c0' or curRxn.id == 'rxn05312_c0' or curRxn.id == 'rxn05315_c0' or curRxn.id == 'rxn10945_c0' or curRxn.id == 'rxn10116_c0':
badRxnList.append(curRxn)
# Transport reactions, where a metabolite with the same ID is on both sides
metabList = []
for curMetab in curRxn.metabolites:
metabList.append(re.sub('_[a-z]\d', '', curMetab.id))
# Count number of appearences each list element
metabList = [metab for metab in metabList if metab != 'cpd00067']
if len(metabList) > 0:
metabCounter = collections.Counter(metabList)
# Sort by number of appearances, find appearances of most common
# Any value > 1 means the metablite is on both sides
if metabCounter.most_common()[0][1] > 1:
badRxnList.append(curRxn)
badRxnList = list(set(badRxnList))
model.remove_reactions(badRxnList, delete=True, remove_orphans=True)
print 'The remaining extracellular metabolites are:'
for curMetab in model.metabolites:
if re.search('_e0', curMetab.id):
print curMetab.id
################################################################################
# Update the metabolite formulas
# print 'Updating metabolite formulas'
for curMetab in model.metabolites:
# Retrieve the metabolite name w/o compartment info
# Look up the appropriate index in the cpdData DF
curMetab.formula = cobra.core.Formula.Formula(cpdData.loc[re.sub(
'_[a-z]\d', '', curMetab.id)][1])
curMetab.formula.id = cpdData.loc[re.sub('_[a-z]\d', '',
curMetab.id)][1]
################################################################################
# Check mass- and charge- balancing
imbalCounter = 0
# print 'Correcting mass- and charge-balancing'
# Check for reactions known to be imbalanced and manually correct them
# If metabolite cpd03422 exists, update its charge to +1
for curMetab in model.metabolites:
if curMetab.id in metabFormDict.keys():
curMetab.formula = cobra.core.Formula.Formula(
metabFormDict[curMetab.id])
curMetab.formula.id = metabFormDict[curMetab.id]
curMetab.charge = int(metabChargeDict[curMetab.id])
for curRxn in model.reactions:
# If reaction rxn05893 exists, update its stoichiometry
if curRxn.id == 'rxn05893_c0':
curRxn.reaction = '4.0 cpd00001_c0 + 2.0 cpd00013_c0 + 6.0 cpd11621_c0 <=> 16.0 cpd00067_c0 + 2.0 cpd00075_c0 + 6.0 cpd11620_c0'
# If reaction rxn07295 exists, update its stoichiometry
if curRxn.id == 'rxn07295_c0':
curRxn.reaction = 'cpd00007_c0 + cpd00033_c0 <=> cpd00025_c0 + 3.0 cpd00067_c0 + cpd14545_c0'
# If reaction rxn08808 exists, update its stoichiometry
elif curRxn.id == 'rxn08808_c0':
curRxn.reaction = 'cpd00001_c0 + cpd15341_c0 <=> cpd00067_c0 + cpd00908_c0 + cpd01080_c0'
# If reaction rxn12822 exists, update its stoichiometry
elif curRxn.id == 'rxn12822_c0':
curRxn.reaction = '2.0 cpd00023_c0 + cpd11621_c0 <=> cpd00024_c0 + cpd00053_c0 + 2.0 cpd00067_c0 + cpd11620_c0'
# If reaction rxn12822 exists, update its stoichiometry
# Heuristic for proton balancing
for curRxn in model.reactions:
imbalDict = curRxn.check_mass_balance()
if len(imbalDict) != 0:
# If imbalancing due to protons alone, correct it
if 'H' in imbalDict and 'charge' in imbalDict and imbalDict[
'H'] == imbalDict['charge']:
curRxn.add_metabolites({
model.metabolites.get_by_id('cpd00067_c0'):
-1 * imbalDict['H']
})
else:
imbalCounter = imbalCounter + 1
print 'Reaction ' + str(curRxn.id) + ' remains unbalanced'
# Inform of results
if imbalCounter != 0:
modelSizeDF.loc[curDir][3] = 0
else:
modelSizeDF.loc[curDir][3] = 1
print 'All reactions are balanced'
################################################################################
### Update names to remove trailing zeros
for curComp in model.compartments:
model.compartments[curComp] = re.sub('_\d', '',
model.compartments[curComp])
model.compartments[re.sub(
'\d', '', curComp)] = model.compartments.pop(curComp)
for curMetab in model.metabolites:
curMetab.id = re.sub('\d$', '', curMetab.id)
curMetab.name = re.sub('_[a-z]\d$', '', curMetab.name)
curMetab.compartment = re.sub('\d$', '', curMetab.compartment)
for curRxn in model.reactions:
curRxn.id = re.sub('\d$', '', curRxn.id)
curRxn.name = re.sub('_[a-z]\d$', '', curRxn.name)
################################################################################
# Store the model properties in the array, write the model output, and increase the counter
modelSizeDF.loc[curDir][0] = len(model.genes)
modelSizeDF.loc[curDir][1] = len(model.metabolites)
modelSizeDF.loc[curDir][2] = len(model.reactions)
# Perform final write to file
# Check that output dir exists and create if necessary
if not os.path.exists('../' + processedDataDir + '/' + curDir):
os.makedirs('../' + processedDataDir + '/' + curDir)
# print 'Writing to file'
cobra.io.write_sbml_model(
model,
'../' + processedDataDir + '/' + curDir + '/' + curDir + '.xml')
count = count + 1
# Write the results to file
modelSizeDF.to_csv('../' + summaryStatsDir + '/modelStats.tsv', sep='\t')
return
def processSBMLforRE(rawModelDir, processedDataDir, summaryStatsDir):
# # Check that folders exist and create them if necessary
if not os.path.exists(processedDataDir):
os.makedirs(processedDataDir)
if not os.path.exists(summaryStatsDir):
os.makedirs(summaryStatsDir)
# Import the list of models
dirList = mf.getDirList(rawModelDir)
numSubDir = len(dirList)
# Import the list of metabolies to revise
metabFormDict = {}
metabChargeDict = {}
with open(dataPath+'/newFormulaDict.txt') as inFile:
for line in inFile:
(key, value) = line.split('\t')
metabFormDict[key] = value
with open(dataPath+'/newChargeDict.txt') as inFile:
for line in inFile:
(key, value) = line.split('\t')
metabChargeDict[key] = value
# Create an array to store results
# Columns: genes, metabs, rxns, balanced (binary)
modelSizeDF = pd.DataFrame(index = dirList, columns=['Genes', 'Metabolites', 'Reactions', 'Balanced'])
# Intialize a counter
count = 1
for curDir in dirList:
# Print the subdirectory name
print 'Processing model ' + curDir + ', ' + str(count) + ' of ' + str(numSubDir)
# Import metabolite charges
cpdData = pd.read_csv(rawModelDir+'/'+curDir+'/'+curDir+'Compounds.tsv', delimiter='\t', index_col=0)
################################################################################
# Before reading in the SBML file, update gene loci so they read in properly
# In the old models ...
# KBase generates gene loci of the form kb|g.######.CDS.###
# Transform them to be of the form curDir_CDS_###
# In the new models ...
# KBase generates gene loci of the form curDir.genome.CDS.###
# Transform them to be of the form curDir_CDS_###
for myLine in fileinput.FileInput(rawModelDir+'/'+curDir+'/'+curDir+'.xml', inplace = True):
# print re.sub('kb\|g\.\d+\.CDS\.(\d+)', curDir+'_CDS_\g<1>', myLine).strip()
print re.sub(curDir+'\.genome\.CDS\.(\d+)', curDir+'_CDS_\g<1>', myLine).strip()
fileinput.close()
################################################################################
# Read in model from SBML
model = cobra.io.read_sbml_model(rawModelDir+'/'+curDir+'/'+curDir+'.xml')
################################################################################
# Remove undesired reactions, including:
badRxnList = []
# print 'Removing bad reactions'
for curRxn in model.reactions:
# Exchange reactions
if re.match('EX_', curRxn.id):
badRxnList.append(curRxn)
# Reactions w/o GPRs
elif len(curRxn.gene_reaction_rule) == 0:
badRxnList.append(curRxn)
# Protein, DNA, and RNA synthesis
elif curRxn.id == 'rxn13782_c0' or curRxn.id == 'rxn13783_c0' or curRxn.id == 'rxn13784_c0':
badRxnList.append(curRxn)
# Spontaneous reactions, whose GPR is fully 'unknown'
elif curRxn.gene_reaction_rule == 'Unknown':
badRxnList.append(curRxn)
# Transport reactions, based on keywords
elif re.search('transport', curRxn.name) or re.search('permease', curRxn.name) or re.search('symport', curRxn.name) or re.search('diffusion', curRxn.name) or re.search('excretion', curRxn.name) or re.search('export', curRxn.name) or re.search('secretion', curRxn.name) or re.search('uptake', curRxn.name) or re.search('antiport', curRxn.name):
badRxnList.append(curRxn)
# Transport reactions which don't get picked up based on keywords
elif curRxn.id == 'rxn05226_c0' or curRxn.id == 'rxn05292_c0' or curRxn.id == 'rxn05305_c0' or curRxn.id == 'rxn05312_c0' or curRxn.id == 'rxn05315_c0' or curRxn.id == 'rxn10945_c0' or curRxn.id == 'rxn10116_c0':
badRxnList.append(curRxn)
# Transport reactions, where a metabolite with the same ID is on both sides
metabList = []
for curMetab in curRxn.metabolites:
metabList.append(re.sub('_[a-z]\d', '', curMetab.id))
# Count number of appearences each list element
metabList = [metab for metab in metabList if metab != 'cpd00067']
if len(metabList) > 0:
metabCounter = collections.Counter(metabList)
# Sort by number of appearances, find appearances of most common
# Any value > 1 means the metablite is on both sides
if metabCounter.most_common()[0][1] > 1:
badRxnList.append(curRxn)
badRxnList = list(set(badRxnList))
model.remove_reactions(badRxnList, delete=True, remove_orphans=True)
print 'The remaining extracellular metabolites are:'
for curMetab in model.metabolites:
if re.search('_e0', curMetab.id):
print curMetab.id
################################################################################
# Update the metabolite formulas
# print 'Updating metabolite formulas'
for curMetab in model.metabolites:
# Retrieve the metabolite name w/o compartment info
# Look up the appropriate index in the cpdData DF
curMetab.formula = cobra.core.Formula.Formula(cpdData.loc[re.sub('_[a-z]\d', '', curMetab.id)][1])
curMetab.formula.id = cpdData.loc[re.sub('_[a-z]\d', '', curMetab.id)][1]
################################################################################
# Check mass- and charge- balancing
imbalCounter = 0
# print 'Correcting mass- and charge-balancing'
# Check for reactions known to be imbalanced and manually correct them
# If metabolite cpd03422 exists, update its charge to +1
for curMetab in model.metabolites:
if curMetab.id in metabFormDict.keys():
curMetab.formula = cobra.core.Formula.Formula(metabFormDict[curMetab.id])
curMetab.formula.id = metabFormDict[curMetab.id]
curMetab.charge = int(metabChargeDict[curMetab.id])
for curRxn in model.reactions:
# If reaction rxn05893 exists, update its stoichiometry
if curRxn.id == 'rxn05893_c0':
curRxn.reaction = '4.0 cpd00001_c0 + 2.0 cpd00013_c0 + 6.0 cpd11621_c0 <=> 16.0 cpd00067_c0 + 2.0 cpd00075_c0 + 6.0 cpd11620_c0'
# If reaction rxn07295 exists, update its stoichiometry
if curRxn.id == 'rxn07295_c0':
curRxn.reaction = 'cpd00007_c0 + cpd00033_c0 <=> cpd00025_c0 + 3.0 cpd00067_c0 + cpd14545_c0'
# If reaction rxn08808 exists, update its stoichiometry
elif curRxn.id == 'rxn08808_c0':
curRxn.reaction = 'cpd00001_c0 + cpd15341_c0 <=> cpd00067_c0 + cpd00908_c0 + cpd01080_c0'
# If reaction rxn12822 exists, update its stoichiometry
elif curRxn.id == 'rxn12822_c0':
curRxn.reaction = '2.0 cpd00023_c0 + cpd11621_c0 <=> cpd00024_c0 + cpd00053_c0 + 2.0 cpd00067_c0 + cpd11620_c0'
# If reaction rxn12822 exists, update its stoichiometry
# Heuristic for proton balancing
for curRxn in model.reactions:
imbalDict = curRxn.check_mass_balance()
if len(imbalDict) != 0:
# If imbalancing due to protons alone, correct it
if 'H' in imbalDict and 'charge' in imbalDict and imbalDict['H'] == imbalDict['charge']:
curRxn.add_metabolites({model.metabolites.get_by_id('cpd00067_c0'): -1*imbalDict['H']})
else:
imbalCounter = imbalCounter + 1
print 'Reaction ' + str(curRxn.id) + ' remains unbalanced'
# Inform of results
if imbalCounter != 0:
modelSizeDF.loc[curDir][3] = 0
else:
modelSizeDF.loc[curDir][3] = 1
print 'All reactions are balanced'
################################################################################
### Update names to remove trailing zeros
for curComp in model.compartments:
model.compartments[curComp] = re.sub('_\d', '', model.compartments[curComp])
model.compartments[re.sub('\d', '', curComp)] = model.compartments.pop(curComp)
for curMetab in model.metabolites:
curMetab.id = re.sub('\d$', '', curMetab.id)
curMetab.name = re.sub('_[a-z]\d$', '', curMetab.name)
curMetab.compartment = re.sub('\d$', '', curMetab.compartment)
for curRxn in model.reactions:
curRxn.id = re.sub('\d$', '', curRxn.id)
curRxn.name = re.sub('_[a-z]\d$', '', curRxn.name)
################################################################################
# Store the model properties in the array, write the model output, and increase the counter
modelSizeDF.loc[curDir][0] = len(model.genes)
modelSizeDF.loc[curDir][1] = len(model.metabolites)
modelSizeDF.loc[curDir][2] = len(model.reactions)
# Perform final write to file
# Check that output dir exists and create if necessary
if not os.path.exists(processedDataDir+'/'+curDir):
os.makedirs(processedDataDir+'/'+curDir)
# print 'Writing to file'
cobra.io.write_sbml_model(model, processedDataDir+'/'+curDir+'/'+curDir+'.xml')
count = count + 1
# Write the results to file
modelSizeDF.to_csv(summaryStatsDir+'/modelStats.tsv', sep='\t')
return
def pruneCurrencyMetabs(modelDir, summaryStatsDir):
# Import the list of models
dirList = mf.getDirList(modelDir)
# Create an array to store results
# Columns: genes, metabs, rxns, balanced (binary)
modelSizeDF = pd.DataFrame(index = dirList, columns=['Genes', 'Metabolites', 'Reactions'])
# Intialize a counter
count = 1
# Process each model...
for curDir in dirList:
# Read in model from SBML
model = cobra.io.read_sbml_model(modelDir+'/'+curDir+'/'+curDir+'.xml')
# Write the original model to a text file for inspection.
# with open(modelDir+'/'+curDir+'/'+curDir+'Original.txt', 'w') as outFile:
# for curRxn in model.reactions:
# outFile.write(curRxn.id+'\t'+curRxn.build_reaction_string(use_metabolite_names=True)+'\n')
#############################
# Read in the list of bad metabolite pairs
pairMetabList = []
with open(dataPath+'/currencyRemovePairs.txt') as myFile:
for line in myFile:
pairMetabList.append(line.strip().split('\t'))
# Remove all pairs of metabolites
for curRxn in model.reactions:
for pair in pairMetabList:
# List metabolites in the reaction. Needs to be inside loop b/c reaction metabolites will change
metabList = []
for curMetab in curRxn.metabolites:
metabList.append(curMetab.id)
# If both members of the pair participate in the reaction, drop both from the reaction
if set(pair) <= set(metabList):
for metab in pair:
curRxn.pop(model.metabolites.get_by_id(metab))
#############################
# Read in the list of aminotransfer metabolite pairs
pairAminoList = []
with open(dataPath+'/currencyAminoPairs.txt') as myFile:
for line in myFile:
pairAminoList.append(line.strip().split('\t'))
# Remove all pairs of metabolites
for curRxn in model.reactions:
for pair in pairAminoList:
# List metabolites in the reaction. Needs to be inside loop b/c reaction metabolites will change
metabList = []
for curMetab in curRxn.metabolites:
metabList.append(curMetab.id)
# If both members of the pair participate in the reaction, drop both from the reaction
if (set(pair) <= set(metabList)) and ('cpd00013_c' not in metabList):
for metab in pair:
curRxn.pop(model.metabolites.get_by_id(metab))
#############################
# Read in the list of bad metabolite singletons
with open(dataPath+'/currencyRemoveSingletons.txt') as myFile:
singletonMetabList = myFile.read().splitlines()
# Remove all bad metabolites
for curMetab in singletonMetabList:
if model.metabolites.has_id(curMetab):
model.metabolites.get_by_id(curMetab).remove_from_model(method='subtractive')
# Prune the model, dropping empty reactions and those with only a subtrate or product
cobra.manipulation.delete.prune_unused_reactions(model)
for curRxn in model.reactions:
if len(curRxn.reactants) == 0 or len(curRxn.products) == 0:
curRxn.remove_from_model(remove_orphans=True)
# Store the model properties in the array, write the model output, and increase the counter
modelSizeDF.loc[curDir][0] = len(model.genes)
modelSizeDF.loc[curDir][1] = len(model.metabolites)
modelSizeDF.loc[curDir][2] = len(model.reactions)
# Write the final model to a text file for inspection
# with open(modelDir+'/'+curDir+'/'+curDir+'Final.txt', 'w') as outFile:
# for curRxn in model.reactions:
# outFile.write(curRxn.id+'\t'+curRxn.build_reaction_string(use_metabolite_names=True)+'\n')
print 'Processing model '+str(count)+' of '+str(len(dirList))
cobra.io.write_sbml_model(model, modelDir+'/'+curDir+'/'+curDir+'.xml')
count = count + 1
# Write the results to file
modelSizeDF.to_csv(summaryStatsDir+'/prunedModelStats.tsv', sep='\t')
return
# These custom-written modules should have been included with the package
# distribution.
import graphFunctions as gf
import metadataFunctions as mf
import sbmlFunctions as sf
import seedFunctions as ef
# Define local folder structure for data input and processing.
processedDataDir = 'ProcessedModelFiles'
summaryStatsDir = 'DataSummaries'
externalDataDir = 'ExternalData'
#%% Program execution
# Import the list of models
dirList = mf.getDirList('../' + processedDataDir)
numSubDir = len(dirList)
# Convert SBML model files to adjacency lists
modelStatArray = sf.dirListToAdjacencyList(dirList, externalDataDir,
processedDataDir, summaryStatsDir)
# Compute statistics on the size of metabolic network graphs
graphStatArray, diGraphStatArray = gf.computeGraphStats(
dirList, processedDataDir, summaryStatsDir)
gf.plotGraphStats(graphStatArray)
# Reduce network graphs to their largest component
reducedGraphStatArray = gf.reduceToLargeComponent(dirList, processedDataDir,
summaryStatsDir)