def computeROC(positivef, negativef):
"""Computes AUC and ROC curve"""
allData = mData.r2Col(positivef)
posSamples = set(allData.keys())
allData = mData.r2Col(negativef, appendData = allData)
negSamples = set(allData.keys()) - posSamples
allSamples = allData.keys()
allSamples.sort(lambda x, y: cmp(allData[y],allData[x]))
x = [0.0]
y = [0.0]
TP = 0
FP = 0
TN = len(negSamples)
FN = len(posSamples)
for i in allSamples:
if i in posSamples:
TP += 1
FN -= 1
else:
TN -= 1
FP += 1
x.append(float(FP)/len(negSamples))
y.append(float(TP)/len(posSamples))
AUC = 0.0
for i in range(len(x)-1):
AUC += (x[i+1]-x[i])*((y[i+1]+y[i])/2)
return(AUC, x, y)
def computeROC(positivef, negativef):
"""Computes AUC and ROC curve"""
allData = mData.r2Col(positivef)
posSamples = set(allData.keys())
allData = mData.r2Col(negativef, appendData=allData)
negSamples = set(allData.keys()) - posSamples
allSamples = allData.keys()
allSamples.sort(lambda x, y: cmp(allData[y], allData[x]))
x = [0.0]
y = [0.0]
TP = 0
FP = 0
TN = len(negSamples)
FN = len(posSamples)
for i in allSamples:
if i in posSamples:
TP += 1
FN -= 1
else:
TN -= 1
FP += 1
x.append(float(FP) / len(negSamples))
y.append(float(TP) / len(posSamples))
AUC = 0.0
for i in range(len(x) - 1):
AUC += (x[i + 1] - x[i]) * ((y[i + 1] + y[i]) / 2)
return (AUC, x, y)
def rSIF(inf, typef = None, reverse = False):
"""read .sif"""
readPathway = Pathway(dict(), dict())
inNodes = dict() #Dictionary with (A : type)
inInteractions = dict() #Dictionary with (A : (B : interaction))
nodeMap = dict()
if typef != None:
nodeMap = mData.r2Col(typef, delim = " = ", header = True)
f = open(inf, "r")
for line in f:
if line.isspace():
continue
line = line.rstrip("\r\n")
pline = re.split("\s*\t\s*", line)
if pline[0] not in inNodes:
if pline[0] in nodeMap:
inNodes[pline[0]] = nodeMap[pline[0]]
else:
inNodes[pline[0]] = "concept"
if pline[2] not in inNodes:
if pline[2] in nodeMap:
inNodes[pline[2]] = nodeMap[pline[2]]
else:
inNodes[pline[2]] = "concept"
if reverse:
if pline[2] not in inInteractions:
inInteractions[pline[2]] = dict()
if pline[0] not in inInteractions[pline[2]]:
inInteractions[pline[2]][pline[0]] = pline[1]
else:
inInteractions[pline[2]][pline[0]] += ";"+pline[1]
else:
if pline[0] not in inInteractions:
inInteractions[pline[0]] = dict()
if pline[2] not in inInteractions[pline[0]]:
inInteractions[pline[0]][pline[2]] = pline[1]
else:
inInteractions[pline[0]][pline[2]] += ";"+pline[1]
f.close()
return(inNodes, inInteractions)
def rSIF(inf, typef=None, reverse=False):
"""read .sif"""
readPathway = Pathway(dict(), dict())
inNodes = dict() #Dictionary with (A : type)
inInteractions = dict() #Dictionary with (A : (B : interaction))
nodeMap = dict()
if typef != None:
nodeMap = mData.r2Col(typef, delim=" = ", header=True)
f = open(inf, "r")
for line in f:
if line.isspace():
continue
line = line.rstrip("\r\n")
pline = re.split("\s*\t\s*", line)
if pline[0] not in inNodes:
if pline[0] in nodeMap:
inNodes[pline[0]] = nodeMap[pline[0]]
else:
inNodes[pline[0]] = "concept"
if pline[2] not in inNodes:
if pline[2] in nodeMap:
inNodes[pline[2]] = nodeMap[pline[2]]
else:
inNodes[pline[2]] = "concept"
if reverse:
if pline[2] not in inInteractions:
inInteractions[pline[2]] = dict()
if pline[0] not in inInteractions[pline[2]]:
inInteractions[pline[2]][pline[0]] = pline[1]
else:
inInteractions[pline[2]][pline[0]] += ";" + pline[1]
else:
if pline[0] not in inInteractions:
inInteractions[pline[0]] = dict()
if pline[2] not in inInteractions[pline[0]]:
inInteractions[pline[0]][pline[2]] = pline[1]
else:
inInteractions[pline[0]][pline[2]] += ";" + pline[1]
f.close()
return (inNodes, inInteractions)
minCol = rgb(255, 0, 0)
log("Color: meth\n")
elif circleFiles[i].startswith("mut."):
maxCol = rgb(0, 0, 0)
minCol = rgb(255, 255, 255)
log("Color: mut\n")
circleColors.append( (minCol, zerCol, maxCol) )
if sampleFile == None:
samples = list(set(cols) | set(samples))
if featureFile == None:
features = list(set(rows) | set(features))
## read centerFile
centerData = None
if centerFile != None:
centerData = mData.r2Col(centerFile, header = True)
## sort
if orderFeature != None:
if len(orderFiles) > 0:
orderData = []
orderColors = []
for i in range(len(orderFiles)):
orderData.append(mData.rCRSData(orderFiles[i]))
minCol = rgb(255, 255, 255)
zerCol = rgb(255, 255, 255)
maxCol = rgb(0, 0, 0)
orderColors.append( (minCol, zerCol, maxCol) )
else:
orderData = circleData
samples.sort(lambda x, y: scmp(x, y, orderFeature, orderData))
def cli_routine(outputDir, circleFiles, orderFiles, sampleFile, featureFile, orderFeature, centerFile, colorscaleFile, printLabel, verbose, cohortMinMax=False, purpleHack = True): """Routine for program execution via command-line.""" # I've tried not to touch this method as much as possible. # I don't want to break the way it was working for Sam Ng. # chrisw ## execute samples = [] features = [] if sampleFile != None: samples = mData.rList(sampleFile) if featureFile != None: features = mData.rList(featureFile) # end section for getting lists of samples and features ## read circleFiles # circleData is a list of dict[col][row]=score from each circleFile circleData = [] # circleColorsPalette is a list of (minColor),(zeroColor),(maxColor) circleColorsPalette = [] ## read colorscaleFile # the format is as follows - header compulsory: # min/max color coding color1 color2 color 3 # -2,2 rgb 155,155,155 255,255,255 0,0,0, # - rgb 155,0,155 255,0,255 0,0,0, # the "color format" is intended to support more color format, as I have # seen the html-colors in the code. # Michael ([email protected]) colorscaleData = None if colorscaleFile != None: if cohortMinMax: log("WARNING: The -k option overrides -m") colorscaleData = mData.retRows(colorscaleFile,aslist=True) line=1 for cs in colorscaleData: line = line + 1 if len(cs) != 5: log("ERROR: color scale needs five fields: datapoints, colorcoding(rgb) and three colors\n", die = True) try: cs[0] = [float(x) for x in cs[0].split(",")] except ValueError: pass if len(cs[0]) != 2 and cs[0] != "-": print cs[0] log("ERROR: Two data points or dash needed for color scale\n", die = True) if cs[1].lower() == "rgb": try: cs[2] = rgb(*[float(x) for x in cs[2].split(",")]) cs[3] = rgb(*[float(x) for x in cs[3].split(",")]) cs[4] = rgb(*[float(x) for x in cs[4].split(",")]) except TypeError: log("ERROR: RGB needs three values on line " + str(line) + "\n", die = True) except ValueError: log("ERROR: RGB color not correctly defined on line " + str(line) + "\n", die=True) else: log("ERROR: Unknown color coding on line " + str(line) + ": " + str(cs[1]) + "\n", die=True) for i in xrange(len(circleFiles)): # get data, samples, and features from each circleFile # data is a dict[col][row]=score # cols is a list of sample names # features is a list of feature names (data, cols, rows) = mData.rCRSData(circleFiles[i], retFeatures=True) circleData.append(data) minCol = lightBlueRGB zerCol = whiteRGB maxCol = redRGB if colorscaleFile != None and i<len(colorscaleData): #get colors from specified colorscaleFile minCol = colorscaleData[i][2] zerCol = colorscaleData[i][3] maxCol = colorscaleData[i][4] # special cases for -meth and -mut # if circleFiles[i].endswith("meth"): # maxCol = blueRGB # minCol = redRGB # log("Color: meth\n") # elif circleFiles[i].endswith("mut"): # maxCol = blackRGB # minCol = whiteRGB # log("Color: mut\n") circleColorsPalette.append((minCol, zerCol, maxCol)) # if no sampleFile/featureFile, default to using samples/features from circleFiles if sampleFile == None: samples = list(set(cols) | set(samples)) if featureFile == None: features = list(set(rows) | set(features)) # end section for reading circleFiles ## read centerFile centerData = None if centerFile != None: centerData = mData.r2Col(centerFile, header=True) ## sort if orderFeature != None: if len(orderFiles) > 0: orderData = [] orderColors = [] for i in xrange(len(orderFiles)): orderData.append(mData.rCRSData(orderFiles[i])) minCol = whiteRGB zerCol = whiteRGB maxCol = blackRGB orderColors.append((minCol, zerCol, maxCol)) else: orderData = circleData # sort samples based on sample score in orderData # priority of sorting determined by orderFiles parameter samples.sort(lambda x, y: scmp(x, y, orderFeature, orderData)) ## cohort png # cgi will probably not use orderFiles if len(orderFiles) > 0: imgFile = "%s/Cohort.png" % (outputDir) label = "Cohort" centerCol = whiteRGB.tohex() cohortCircleCols = [] for i in xrange(len(orderData)): ringCols = [] ringVals = [] for sample in samples: if sample in orderData[i]: if orderFeature in orderData[i][sample]: ringVals.append(orderData[i][sample][orderFeature]) elif "*" in orderData[i][sample]: ringVals.append(orderData[i][sample]["*"]) minVal = min([-0.01] + mData.floatList(ringVals)) maxVal = max([0.01] + mData.floatList(ringVals)) for sample in samples: if sample in orderData[i]: if orderFeature in orderData[i][sample]: ringCols.append(getColor(orderData[i][sample][orderFeature], minVal, maxVal, minColor=orderColors[i][0], zeroColor=orderColors[i][1], maxColor=orderColors[i][2])) elif "*" in orderData[i][sample]: ringCols.append(getColor(orderData[i][sample]["*"], minVal, maxVal, minColor=orderColors[i][0], zeroColor=orderColors[i][1], maxColor=orderColors[i][2])) else: ringCols.append(greyRGB.tohex()) else: ringCols.append(greyRGB.tohex()) cohortCircleCols.append(ringCols) plotCircle(imgFile, label=label, centerCol=centerCol, circleCols=cohortCircleCols, innerRadTotal=0.2, outerRadTotal=0.5, width=5) # end section for sample ordering ## plot images if centerData != None: centerDataFloatList = mData.floatList(centerData.values()) centerDataMinVal = min([-0.01] + centerDataFloatList) centerDataMaxVal = max([0.01] + centerDataFloatList) # get min/max values for datasets if cohortMinMax: (minValList, maxValList) = getCohortMinMaxValues(features, samples, circleData) else: (minValList, maxValList) = (None, None) if colorscaleData != None: (minValList, maxValList) = getColorScaleMinMaxValues(minValList, maxValList, len(circleData), colorscaleData) for feature in features: log("Drawing %s\n" % (feature)) centerColHex = None if centerData != None: if feature in centerData: centerColHex = getColor(centerData[feature], centerDataMinVal, centerDataMaxVal, minColor=lightBlueRGB, zeroColor=whiteRGB, purple0Hack=purpleHack) imgFile = "%s/%s.png" % (outputDir, re.sub("[/:]", "_", feature)) label = "" if printLabel: label = feature image_width = 5.0 drawCircleImageForFeature(feature, samples, label, imgFile, circleData, circleColorsPalette, width=image_width, centerColHex=centerColHex, minValList=minValList, maxValList=maxValList, purple0Hack=purpleHack) for sample in samples: log("ordered samples: %s\n" % (sample))
if os.path.exists("%s/img" % (feature)):
customImage = True
## identify nets with feature
sifFile = None
for i in os.listdir("%s" % (feature+"/")):
if i.endswith(netExtension):
sifFile = feature+"/"+i
break
(allNodes, allInteractions) = mPathway.rSIF(sifFile)
idMap = dict()
nodeMap = dict()
for i, node in enumerate(allNodes.keys()):
idMap[node] = i+1
nodeMap[i+1] = node
labelMap = mData.r2Col("LABEL.NA", delim = " = ", header = True)
typeMap = mData.r2Col("TYPE.NA", delim = " = ", header = True)
scoreMap = mData.r2Col("%s_SCORE.NA" % (feature), delim = " = ", header = True)
nodes = nodeMap.keys()
nodes.sort()
## create graphml structure
graphmlContent = """<graphml>\\
<key id="name" for="node" attr.name="name" attr.type="string"/>\\
<key id="label" for="node" attr.name="label" attr.type="string"/>\\
<key id="type" for="node" attr.name="type" attr.type="string"/>\\
<key id="color" for="node" attr.name="color" attr.type="string"/>\\
<key id="size" for="node" attr.name="size" attr.type="double"/>\\
<key id="score" for="node" attr.name="score" attr.type="double"/>\\
<key id="image" for="node" attr.name="image" attr.type="string"/>\\
<key id="interaction" for="edge" attr.name="interaction" attr.type="string"/>\\
def cli_routine(outputDir,
circleFiles,
orderFiles,
sampleFile,
featureFile,
orderFeature,
centerFile,
colorscaleFile,
printLabel,
verbose,
cohortMinMax=False,
purpleHack=True):
"""Routine for program execution via command-line."""
# I've tried not to touch this method as much as possible.
# I don't want to break the way it was working for Sam Ng.
# chrisw
## execute
samples = []
features = []
if sampleFile != None:
samples = mData.rList(sampleFile)
if featureFile != None:
features = mData.rList(featureFile)
# end section for getting lists of samples and features
## read circleFiles
# circleData is a list of dict[col][row]=score from each circleFile
circleData = []
# circleColorsPalette is a list of (minColor),(zeroColor),(maxColor)
circleColorsPalette = []
## read colorscaleFile
# the format is as follows - header compulsory:
# min/max color coding color1 color2 color 3
# -2,2 rgb 155,155,155 255,255,255 0,0,0,
# - rgb 155,0,155 255,0,255 0,0,0,
# the "color format" is intended to support more color format, as I have
# seen the html-colors in the code.
# Michael ([email protected])
colorscaleData = None
if colorscaleFile != None:
if cohortMinMax:
log("WARNING: The -k option overrides -m")
colorscaleData = mData.retRows(colorscaleFile, aslist=True)
line = 1
for cs in colorscaleData:
line = line + 1
if len(cs) != 5:
log("ERROR: color scale needs five fields: datapoints, colorcoding(rgb) and three colors\n",
die=True)
try:
cs[0] = [float(x) for x in cs[0].split(",")]
except ValueError:
pass
if len(cs[0]) != 2 and cs[0] != "-":
print cs[0]
log("ERROR: Two data points or dash needed for color scale\n",
die=True)
if cs[1].lower() == "rgb":
try:
cs[2] = rgb(*[float(x) for x in cs[2].split(",")])
cs[3] = rgb(*[float(x) for x in cs[3].split(",")])
cs[4] = rgb(*[float(x) for x in cs[4].split(",")])
except TypeError:
log("ERROR: RGB needs three values on line " + str(line) +
"\n",
die=True)
except ValueError:
log("ERROR: RGB color not correctly defined on line " +
str(line) + "\n",
die=True)
else:
log("ERROR: Unknown color coding on line " + str(line) + ": " +
str(cs[1]) + "\n",
die=True)
for i in xrange(len(circleFiles)):
# get data, samples, and features from each circleFile
# data is a dict[col][row]=score
# cols is a list of sample names
# features is a list of feature names
(data, cols, rows) = mData.rCRSData(circleFiles[i], retFeatures=True)
circleData.append(data)
minCol = lightBlueRGB
zerCol = whiteRGB
maxCol = redRGB
if colorscaleFile != None and i < len(colorscaleData):
#get colors from specified colorscaleFile
minCol = colorscaleData[i][2]
zerCol = colorscaleData[i][3]
maxCol = colorscaleData[i][4]
# special cases for -meth and -mut
# if circleFiles[i].endswith("meth"):
# maxCol = blueRGB
# minCol = redRGB
# log("Color: meth\n")
# elif circleFiles[i].endswith("mut"):
# maxCol = blackRGB
# minCol = whiteRGB
# log("Color: mut\n")
circleColorsPalette.append((minCol, zerCol, maxCol))
# if no sampleFile/featureFile, default to using samples/features from circleFiles
if sampleFile == None:
samples = list(set(cols) | set(samples))
if featureFile == None:
features = list(set(rows) | set(features))
# end section for reading circleFiles
## read centerFile
centerData = None
if centerFile != None:
centerData = mData.r2Col(centerFile, header=True)
## sort
if orderFeature != None:
if len(orderFiles) > 0:
orderData = []
orderColors = []
for i in xrange(len(orderFiles)):
orderData.append(mData.rCRSData(orderFiles[i]))
minCol = whiteRGB
zerCol = whiteRGB
maxCol = blackRGB
orderColors.append((minCol, zerCol, maxCol))
else:
orderData = circleData
# sort samples based on sample score in orderData
# priority of sorting determined by orderFiles parameter
samples.sort(lambda x, y: scmp(x, y, orderFeature, orderData))
## cohort png
# cgi will probably not use orderFiles
if len(orderFiles) > 0:
imgFile = "%s/Cohort.png" % (outputDir)
label = "Cohort"
centerCol = whiteRGB.tohex()
cohortCircleCols = []
for i in xrange(len(orderData)):
ringCols = []
ringVals = []
for sample in samples:
if sample in orderData[i]:
if orderFeature in orderData[i][sample]:
ringVals.append(orderData[i][sample][orderFeature])
elif "*" in orderData[i][sample]:
ringVals.append(orderData[i][sample]["*"])
minVal = min([-0.01] + mData.floatList(ringVals))
maxVal = max([0.01] + mData.floatList(ringVals))
for sample in samples:
if sample in orderData[i]:
if orderFeature in orderData[i][sample]:
ringCols.append(
getColor(orderData[i][sample][orderFeature],
minVal,
maxVal,
minColor=orderColors[i][0],
zeroColor=orderColors[i][1],
maxColor=orderColors[i][2]))
elif "*" in orderData[i][sample]:
ringCols.append(
getColor(orderData[i][sample]["*"],
minVal,
maxVal,
minColor=orderColors[i][0],
zeroColor=orderColors[i][1],
maxColor=orderColors[i][2]))
else:
ringCols.append(greyRGB.tohex())
else:
ringCols.append(greyRGB.tohex())
cohortCircleCols.append(ringCols)
plotCircle(imgFile,
label=label,
centerCol=centerCol,
circleCols=cohortCircleCols,
innerRadTotal=0.2,
outerRadTotal=0.5,
width=5)
# end section for sample ordering
## plot images
if centerData != None:
centerDataFloatList = mData.floatList(centerData.values())
centerDataMinVal = min([-0.01] + centerDataFloatList)
centerDataMaxVal = max([0.01] + centerDataFloatList)
# get min/max values for datasets
if cohortMinMax:
(minValList,
maxValList) = getCohortMinMaxValues(features, samples, circleData)
else:
(minValList, maxValList) = (None, None)
if colorscaleData != None:
(minValList,
maxValList) = getColorScaleMinMaxValues(minValList, maxValList,
len(circleData),
colorscaleData)
for feature in features:
log("Drawing %s\n" % (feature))
centerColHex = None
if centerData != None:
if feature in centerData:
centerColHex = getColor(centerData[feature],
centerDataMinVal,
centerDataMaxVal,
minColor=lightBlueRGB,
zeroColor=whiteRGB,
purple0Hack=purpleHack)
imgFile = "%s/%s.png" % (outputDir, re.sub("[/:]", "_", feature))
label = ""
if printLabel:
label = feature
image_width = 5.0
drawCircleImageForFeature(feature,
samples,
label,
imgFile,
circleData,
circleColorsPalette,
width=image_width,
centerColHex=centerColHex,
minValList=minValList,
maxValList=maxValList,
purple0Hack=purpleHack)
for sample in samples:
log("ordered samples: %s\n" % (sample))
if os.path.exists("%s/img" % (feature)):
customImage = True
## identify nets with feature
sifFile = None
for i in os.listdir("%s" % (feature + "/")):
if i.endswith(netExtension):
sifFile = feature + "/" + i
break
(allNodes, allInteractions) = mPathway.rSIF(sifFile)
idMap = dict()
nodeMap = dict()
for i, node in enumerate(allNodes.keys()):
idMap[node] = i + 1
nodeMap[i + 1] = node
labelMap = mData.r2Col("LABEL.NA", delim=" = ", header=True)
typeMap = mData.r2Col("TYPE.NA", delim=" = ", header=True)
scoreMap = mData.r2Col("%s_SCORE.NA" % (feature), delim=" = ", header=True)
nodes = nodeMap.keys()
nodes.sort()
## create graphml structure
graphmlContent = """<graphml>\\
<key id="name" for="node" attr.name="name" attr.type="string"/>\\
<key id="label" for="node" attr.name="label" attr.type="string"/>\\
<key id="type" for="node" attr.name="type" attr.type="string"/>\\
<key id="color" for="node" attr.name="color" attr.type="string"/>\\
<key id="size" for="node" attr.name="size" attr.type="double"/>\\
<key id="score" for="node" attr.name="score" attr.type="double"/>\\
<key id="image" for="node" attr.name="image" attr.type="string"/>\\
<key id="interaction" for="edge" attr.name="interaction" attr.type="string"/>\\
log("ERROR: incorrect number of arguments", die = True)
featureFile = args[0]
pathwayFile = args[1]
scoreFile = args[2]
global verbose
for o, a in opts:
if o == "-q":
verbose = False
## execute
featureList = mData.rList(featureFile)
(gNodes, gInteractions) = mPathway.rPathway(pathwayFile)
scoreMap = {}
scoreMap[sessionName] = mData.r2Col(scoreFile)
## find connected
connectList = set()
for source in featureList:
if source not in gNodes:
continue
for target in featureList:
if target not in gNodes:
continue
if source == target:
continue
paths = mPathway.shortestPath(source, target, gInteractions, maxDistance = maxDistance)
if len(paths) == 0:
log("%s /> %s\n" % (source, target))
else:
