def performNMF(species, NMFinput, full_PSI_InputFile, filtered_EventAnnot_dir, k, AnalysisRound, strategy):
""" Run NMF and determine the number of valid clusters based on the magnitude of detected differential splicing """
use_adjusted_p=True
print "Running NMF analyses for dimension reduction using "+str(k)+" k - Round"+str(AnalysisRound)
NMFResult,BinarizedOutput,metaData,Annotation=NMF_Analysis.NMFAnalysis(NMFinput,k,AnalysisRound,strategy) ### This is where we get the correct version
print "Running metaData Analyses for finding differential splicing events"
rootdir,CovariateQuery=metaDataAnalysis.remoteAnalysis(species,filtered_EventAnnot_dir,metaData,'PSI',0.1,use_adjusted_p,0.05,Annotation)
counter=1
dPSI_results_dir=rootdir+CovariateQuery
global upd_guides
upd_guides=[]
name=[]
group=[]
grplst=[]
for filename in os.listdir(dPSI_results_dir):
if filename.startswith("PSI."):
dPSI_results_fn=os.path.join(dPSI_results_dir, filename)
dPSI_comparison_alt_name=string.replace(filename,"PSI.","")
omitcluster=FindTopUniqueEvents(dPSI_results_fn,dPSI_comparison_alt_name,dPSI_results_dir)
if omitcluster==0: ### Hence, clustering succeeded and did not fail in this dPSI comparison
group.append(counter)
name.append(string.replace(filename,"PSI.",""))
counter+=1
print counter, 'robust splicing subtypes identified in round',AnalysisRound
if counter>0: #counter>2 --- changed to 0 to force NMF
dire = export.findParentDir(full_PSI_InputFile)
output_dir = dire+'OncoInputs'
if os.path.exists(output_dir)==False:
export.createExportFolder(output_dir)
output_file = output_dir+'/SVMInput-Round'+str(AnalysisRound)+'.txt'
ExpandSampleClusters.filterRows(full_PSI_InputFile,output_file,filterDB=upd_guides,logData=False)
header=ExpandSampleClusters.header_file(output_file)
print "Running SVM prediction for improved subtypes - Round"+str(AnalysisRound)
#print 'AAAAAAAAAAAAAAAAAAAAAAAA',output_file
#print 'BBBBBBBBBBBBBBBBBBBBBBBB',BinarizedOutput
train=ExpandSampleClusters.TrainDataGeneration(output_file,BinarizedOutput,name)
grplst.append(group)
ExpandSampleClusters.Classify(header,train,output_file,grplst,name,AnalysisRound) ### This is where we write the worng version
header=Correlationdepletion.header_file(NMFResult)
output_file=output_dir+'/DepletionInput-Round'+str(AnalysisRound)+".txt"
sampleIndexSelection.filterFile(full_PSI_InputFile,output_file,header)
print "Running Correlation Depletion - Round"+str(AnalysisRound)
commonkeys,count=Correlationdepletion.FindCorrelations(NMFResult,output_file,name)
Depleted=Correlationdepletion.DepleteSplicingevents(commonkeys,output_file,count,full_PSI_InputFile)
full_PSI_InputFile=Depleted
flag=True ### Indicates that K-means was not run - hence, another round of splice-ICGS should be performed
""""
else:
try:
print "Running K-means analyses instead of NMF - Round"+str(AnalysisRound)
header=[]
header=Kmeans.header_file(dPSI_results_fn_block)
Kmeans.KmeansAnalysis(dPSI_results_fn_block,header,full_PSI_InputFile,AnalysisRound)
flag=True
except Exception:
print 'WARNING!!!!! DID NOT RUN K-MEANS!!!!!'
print traceback.format_exc()
AnalysisRound = True
"""
return flag,full_PSI_InputFile
if EnrichmentOnly==False:
print 'PSI input files:',EventAnnot
print 'Using a rho-cutoff of:',rho_cutoff
if filters==True: ### Filter based on a default percentage of samples with detected PSI values
EventAnnot,SampleNumber=filterPSIValues(EventAnnot,percentCutoff=percentCutoff,filterStatus=True)
else:
SampleNumber=filterPSIValues(EventAnnot,percentCutoff=percentCutoff,filterStatus=False)
output_dir = dire+'ExpressionInput'
export.createExportFolder(output_dir)
full_PSI_InputFile=output_dir+"/exp.input.txt"
header=header_list(EventAnnot)
sampleIndexSelection.filterFile(EventAnnot,full_PSI_InputFile,header,FirstCol=False)
### Set Splice-ICGS defaults
gsp = UI.GeneSelectionParameters(species,platform,platform)
gsp.setNormalize('median')
gsp.setGeneSelection('')
gsp.setGeneSet('None Selected')
gsp.setPathwaySelect([])
gsp.setJustShowTheseIDs('')
gsp.setSampleDiscoveryParameters(ExpressionCutoff,CountsCutoff,FoldDiff,SamplesDiffering,removeOutliers,
featurestoEvaluate,restrictBy,excludeCellCycle,column_metric,column_method,rho_cutoff)
AnalysisRound=1
if mode == "single":
""" Perform a single round of Splice-ICGS (RNASeq.py module) """
flag,full_PSI_InputFile,EventAnnot=CompleteWorkflow(full_PSI_InputFile,EventAnnot,rho_cutoff,strategy,seq,gsp,forceBroadClusters,AnalysisRound)
if __name__ == '__main__':
import getopt
################ Comand-line arguments ################
if len(
sys.argv[1:]
) <= 1: ### Indicates that there are insufficient number of command-line arguments
print "Warning! Insufficient command line flags supplied."
sys.exit()
else:
analysisType = []
options, remainder = getopt.getopt(sys.argv[1:], '',
['Guidefile=', 'PSIfile='])
for opt, arg in options:
if opt == '--Guidefile': Guidefile = arg
elif opt == '--PSIfile': PSIfile = arg
else:
print "Warning! Command-line argument: %s not recognized. Exiting..." % opt
sys.exit()
#filename="/Users/meenakshi/Documents/leucegene/ICGS/Clustering-exp.Hs_RNASeq_top_alt_junctions367-Leucegene-75p_no149-Guide1 TRAK1&ENSG00000182606&I1.1_42075542-E2.1__E-hierarchical_cosine_correlation.txt"
#PSIfile="/Users/meenakshi/Documents/leucegene/ExpressionInput/exp.Hs_RNASeq_top_alt_junctions-PSI_EventAnnotation-367-Leucegene-75p-unique-filtered-filtered.txt"
#keylabel="/Users/meenakshi/Documents/leucegene/ExpressionInput/exp.round2_glmfilteredKmeans_label.txt"
header = header_file(Guidefile)
output_file = PSIfile[:-4] + "-filtered.txt"
sampleIndexSelection.filterFile(PSIfile, output_file, header)
commonkeys, count = FindCorrelations(Guidefile, output_file)
DepleteSplicingevents(commonkeys, output_file, count)
def CompleteWorkflow(InputFile, EventAnnot, rho_cutoff, strategy, seq, gsp,
forceBroadClusters, turn):
""" This function is used perform a single-iteration of the OncoSplice workflow (called from main),
including the unsupervised splicing analysis (splice-ICGS) and signature depletion """
### Filter the EventAnnotation PSI file with non-depleted events from the prior round
FilteredEventAnnot = filterEventAnnotation.FilterFile(
InputFile, EventAnnot, turn)
try:
print "Running splice-ICGS for feature selection - Round" + str(turn)
### Reset the below variables which can be altered in prior rounds
gsp.setGeneSelection('')
gsp.setGeneSet('None Selected')
gsp.setPathwaySelect([])
if forceBroadClusters == True:
### Find Broad clusters with at least 25% of all samples
originalSamplesDiffering = gsp.SamplesDiffering()
gsp.setSamplesDiffering(int(SampleNumber * 0.25))
print 'Number varying samples to identify:', gsp.SamplesDiffering()
graphic_links3 = RNASeq.singleCellRNASeqWorkflow(species,
'exons',
InputFile,
mlp,
exp_threshold=0,
rpkm_threshold=0,
parameters=gsp)
if forceBroadClusters == True:
gsp.setSamplesDiffering(originalSamplesDiffering)
Guidefile = graphic_links3[-1][-1]
Guidefile = Guidefile[:-4] + '.txt'
print "Running block identification for rank analyses - Round" + str(
turn)
### Parameters are fixed as they are distinct
RNASeq_blockIdentification.correlateClusteredGenesParameters(
Guidefile,
rho_cutoff=0.4,
hits_cutoff=4,
hits_to_report=50,
ReDefinedClusterBlocks=True,
filter=True)
Guidefile_block = Guidefile[:-4] + '-BlockIDs.txt'
NMFinput, Rank = NMF_Analysis.FilterFile(Guidefile, Guidefile_block,
InputFile, turn)
except Exception:
print 'UNKNOWN ERROR!!!!! Setting Rank=0'
#print traceback.format_exc()
Rank = 0
if Rank > 1:
### ADJUST THE RANKS - MUST UPDATE!!!!
if turn == 1:
if force_broad_round1:
#Rank=2
Rank = Rank
else:
if Rank > 2:
Rank = 30
else:
if Rank > 2:
Rank = 30
if seq == "bulk":
use_adjusted_p = True
else:
use_adjusted_p = False
print "Running NMF analyses for dimension reduction using " + str(
Rank) + " ranks - Round" + str(turn)
NMFResult, BinarizedOutput, Metadata, Annotation = NMF_Analysis.NMFAnalysis(
NMFinput, Rank, turn, strategy)
print "Running Metadata Analyses for finding differential splicing events"
rootdir, CovariateQuery = metaDataAnalysis.remoteAnalysis(
'Hs', FilteredEventAnnot, Metadata, 'PSI', 0.1, use_adjusted_p,
0.05, Annotation)
counter = 1
Guidedir = rootdir + CovariateQuery
PSIdir = rootdir + 'ExpressionProfiles'
global upd_guides
upd_guides = []
name = []
group = []
grplst = []
for filename in os.listdir(Guidedir):
if filename.startswith("PSI."):
Guidefile = os.path.join(Guidedir, filename)
psi = string.replace(filename, "PSI.", "")
PSIfile = os.path.join(PSIdir, psi)
omitcluster = FindTopUniqueEvents(Guidefile, psi, Guidedir)
if omitcluster == 0:
group.append(counter)
name.append(psi)
counter += 1
if counter > 2:
dire = export.findParentDir(InputFile)
output_dir = dire + 'OncoInputs'
if os.path.exists(output_dir) == False:
export.createExportFolder(output_dir)
output_file = output_dir + '/SVMInput-Round' + str(turn) + '.txt'
ExpandSampleClusters.filterRows(InputFile,
output_file,
filterDB=upd_guides,
logData=False)
header = ExpandSampleClusters.header_file(output_file)
print "Running SVM prediction for improved subtypes - Round" + str(
turn)
train = ExpandSampleClusters.TrainDataGeneration(
output_file, BinarizedOutput, name)
grplst.append(group)
ExpandSampleClusters.Classify(header, train, output_file, grplst,
name, turn)
header = Correlationdepletion.header_file(NMFResult)
output_file = output_dir + '/DepletionInput-Round' + str(
turn) + ".txt"
sampleIndexSelection.filterFile(InputFile, output_file, header)
print "Running Correlation Depletion - Round" + str(turn)
commonkeys, count = Correlationdepletion.FindCorrelations(
NMFResult, output_file, name)
Depleted = Correlationdepletion.DepleteSplicingevents(
commonkeys, output_file, count, InputFile)
InputFile = Depleted
flag = True
else:
try:
print "Running K-means analyses instead of NMF - Round" + str(
turn)
header = []
header = Kmeans.header_file(Guidefile_block)
Kmeans.KmeansAnalysis(Guidefile_block, header, InputFile, turn)
flag = False
except Exception:
print 'WARNING!!!!! DID NOT RUN K-MEANS!!!!!'
print traceback.format_exc()
flag = False
else:
if Rank == 1:
try:
print "Running K-means analyses instead of NMF - Round" + str(
turn)
header = []
header = Kmeans.header_file(Guidefile_block)
Kmeans.KmeansAnalysis(Guidefile_block, header, InputFile, turn)
flag = False
except Exception:
print 'WARNING!!!!! DID NOT RUN K-MEANS!!!!!'
print traceback.format_exc()
flag = False
else:
flag = False
return flag, InputFile, FilteredEventAnnot
def CompleteWorkflow(InputFile, EventAnnot, turn, rho_cutoff, strategy, seq):
species = "Hs"
row_method = 'hopach'
column_method = 'hopach'
row_metric = 'correlation'
column_metric = 'euclidean'
color_gradient = 'yellow_black_blue'
contrast = 3
vendor = "RNASeq"
GeneSelection = ''
PathwaySelection = ''
GeneSetSelection = 'None Selected'
excludeCellCycle = False
#rho_cutoff = 0.4
restrictBy = 'protein_coding'
featurestoEvaluate = 'Genes'
ExpressionCutoff = 0
CountsCutoff = 0
FoldDiff = 1.2
SamplesDiffering = 4
JustShowTheseIDs = ''
removeOutliers = False
PathwaySelection = []
array_type = "RNASeq"
#rho_cutoff=0.4
gsp = UI.GeneSelectionParameters(species, array_type, vendor)
gsp.setGeneSet(GeneSetSelection)
gsp.setPathwaySelect(PathwaySelection)
gsp.setGeneSelection(GeneSelection)
gsp.setJustShowTheseIDs(JustShowTheseIDs)
gsp.setNormalize('median')
gsp.setSampleDiscoveryParameters(ExpressionCutoff, CountsCutoff, FoldDiff,
SamplesDiffering, removeOutliers,
featurestoEvaluate, restrictBy,
excludeCellCycle, column_metric,
column_method, rho_cutoff)
#Run splice ICGS
"""import UI
species='Mm'; platform = "3'array"; vendor = 'Ensembl'
gsp = UI.GeneSelectionParameters(species,platform,vendor)
gsp.setGeneSet('None Selected')
gsp.setPathwaySelect('')
gsp.setGeneSelection('')
gsp.setJustShowTheseIDs('')
gsp.setNormalize('median')
gsp.setSampleDiscoveryParameters(0,0,1.5,3,
False,'PSI','protein_coding',False,'cosine','hopach',0.35)"""
FilteredEventAnnot = filterEventAnnotation.FilterFile(
InputFile, EventAnnot, turn)
try:
print "Running splice-ICGS for feature selection - Round" + str(turn)
#except Exception:Rank=0
graphic_links3 = RNASeq.singleCellRNASeqWorkflow(species,
'exons',
InputFile,
mlp,
exp_threshold=0,
rpkm_threshold=0,
parameters=gsp)
Guidefile = graphic_links3[-1][-1]
Guidefile = Guidefile[:-4] + '.txt'
print "Running block identification for rank analyses - Round" + str(
turn)
RNASeq_blockIdentification.correlateClusteredGenesParameters(
Guidefile,
rho_cutoff=0.4,
hits_cutoff=4,
hits_to_report=50,
ReDefinedClusterBlocks=True,
filter=True)
Guidefile_block = Guidefile[:-4] + '-BlockIDs.txt'
NMFinput, Rank = NMF_Analysis.FilterFile(Guidefile, Guidefile_block,
InputFile, turn)
except Exception:
print 'UNKNOWN ERROR!!!!!'
print traceback.format_exc()
Rank = 0
if Rank > 1:
print 'Current turn:', turn, 'k =',
if turn == 1:
Rank = 2
elif Rank > 2:
Rank = 30
else:
Rank = 2
if seq == "bulk":
use_adjusted_p = True
else:
use_adjusted_p = False
print Rank
print "Running NMF analyses for dimension reduction using " + str(
Rank) + " ranks - Round" + str(turn)
NMFResult, BinarizedOutput, Metadata, Annotation = NMF_Analysis.NMFAnalysis(
NMFinput, Rank, turn, strategy)
print "Running Metadata Analyses for finding differential splicing events"
rootdir, CovariateQuery = metaDataAnalysis.remoteAnalysis(
'Hs', FilteredEventAnnot, Metadata, 'PSI', 0.1, use_adjusted_p,
0.05, Annotation)
counter = 1
Guidedir = rootdir + CovariateQuery
PSIdir = rootdir + 'ExpressionProfiles'
global upd_guides
upd_guides = []
name = []
group = []
grplst = []
for filename in os.listdir(Guidedir):
if filename.startswith("PSI."):
Guidefile = os.path.join(Guidedir, filename)
psi = string.replace(filename, "PSI.", "")
PSIfile = os.path.join(PSIdir, psi)
omitcluster = FindTopUniqueEvents(Guidefile, psi, Guidedir)
if omitcluster == 0:
group.append(counter)
name.append(psi)
counter += 1
if counter > 2:
dire = export.findParentDir(InputFile)
output_dir = dire + 'OncoInputs'
if os.path.exists(output_dir) == False:
export.createExportFolder(output_dir)
output_file = output_dir + '/SVMInput-Round' + str(turn) + '.txt'
ExpandSampleClusters.filterRows(InputFile,
output_file,
filterDB=upd_guides,
logData=False)
header = ExpandSampleClusters.header_file(output_file)
print "Running SVM prediction for improved subtypes - Round" + str(
turn)
train = ExpandSampleClusters.TrainDataGeneration(
output_file, BinarizedOutput, name)
grplst.append(group)
ExpandSampleClusters.Classify(header, train, output_file, grplst,
name, turn)
header = Correlationdepletion.header_file(NMFResult)
output_file = output_dir + '/DepletionInput-Round' + str(
turn) + ".txt"
sampleIndexSelection.filterFile(InputFile, output_file, header)
print "Running Correlation Depletion - Round" + str(turn)
commonkeys, count = Correlationdepletion.FindCorrelations(
NMFResult, output_file, name)
Depleted = Correlationdepletion.DepleteSplicingevents(
commonkeys, output_file, count, InputFile)
InputFile = Depleted
flag = True
else:
try:
print "Running K-means analyses instead of NMF - Round" + str(
turn)
header = []
header = Kmeans.header_file(Guidefile_block)
Kmeans.KmeansAnalysis(Guidefile_block, header, InputFile, turn)
flag = False
except Exception:
print 'WARNING!!!!! DID NOT RUN K-MEANS!!!!!'
print traceback.format_exc()
flag = False
else:
if Rank == 1:
try:
print "Running K-means analyses instead of NMF - Round" + str(
turn)
header = []
header = Kmeans.header_file(Guidefile_block)
Kmeans.KmeansAnalysis(Guidefile_block, header, InputFile, turn)
flag = False
except Exception:
print 'WARNING!!!!! DID NOT RUN K-MEANS!!!!!'
print traceback.format_exc()
flag = False
else:
flag = False
return flag, InputFile, FilteredEventAnnot