def ClCla(alignedstack, numpart=None, numclasses=40,
factorlist=range(1,5), corandata="coran/corandata", dataext=".spi"):
"""
this doesn't work
"""
if alignedstack[-4:] == dataext:
alignedstack = alignedstack[:-4]
rundir = "cluster"
classavg = rundir+"/"+("classavgstack%03d" % numclasses)
classvar = rundir+"/"+("classvarstack%03d" % numclasses)
apParam.createDirectory(rundir)
for i in range(numclasses):
apFile.removeFile(rundir+("/classdoc%04d" % (i+1))+dataext)
apFile.removeFile(rundir+"/clusterdoc"+dataext)
factorstr, factorkey = operations.intListToString(factorlist)
### do hierarchical clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=True)
mySpider.toSpider(
"CL CLA",
corandata, # path to coran data
rundir+"/clusterdoc", #clusterdoc file
factorstr, #factor numbers
"5,8",
"4",
"2", # minimum number of particles per class
"Y", rundir+"/dendrogram.ps",
"Y", rundir+"/dendrogramdoc",
)
mySpider.close()
def ClCla(alignedstack, numpart=None, numclasses=40,
factorlist=range(1,5), corandata="coran/corandata", dataext=".spi"):
"""
this doesn't work
"""
if alignedstack[-4:] == dataext:
alignedstack = alignedstack[:-4]
rundir = "cluster"
classavg = rundir+"/"+("classavgstack%03d" % numclasses)
classvar = rundir+"/"+("classvarstack%03d" % numclasses)
apParam.createDirectory(rundir)
for i in range(numclasses):
apFile.removeFile(rundir+("/classdoc%04d" % (i+1))+dataext)
apFile.removeFile(rundir+"/clusterdoc"+dataext)
factorstr, factorkey = operations.intListToString(factorlist)
### do hierarchical clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=True)
mySpider.toSpider(
"CL CLA",
corandata, # path to coran data
rundir+"/clusterdoc", #clusterdoc file
factorstr, #factor numbers
"5,8",
"4",
"2", # minimum number of particles per class
"Y", rundir+"/dendrogram.ps",
"Y", rundir+"/dendrogramdoc",
)
mySpider.close()
def setRunDir(self): stackdata = apStack.getOnlyStackData(self.params['tiltstackid'], msg=False) path = stackdata['path']['path'] uppath = os.path.dirname(os.path.dirname(os.path.abspath(path))) classliststr = operations.intListToString(self.classlist) self.params['rundir'] = os.path.join(uppath, "rctvolume", self.params['runname'] )
def setRunDir(self):
stackdata = apStack.getOnlyStackData(self.params['tiltstackid'],
msg=False)
path = stackdata['path']['path']
uppath = os.path.dirname(os.path.dirname(os.path.abspath(path)))
classliststr = operations.intListToString(self.classlist)
self.params['rundir'] = os.path.join(uppath, "rctvolume",
self.params['runname'])
def insertRctRun(self, volfile): ### setup resolutions fscresq = appiondata.ApResolutionData() fscresq['type'] = "fsc" fscresq['half'] = self.fscresolution fscresq['fscfile'] = "fscdata"+self.timestamp+".fsc" rmeasureq = appiondata.ApResolutionData() rmeasureq['type'] = "rmeasure" rmeasureq['half'] = self.rmeasureresolution rmeasureq['fscfile'] = None ### insert rct run data rctrunq = appiondata.ApRctRunData() rctrunq['runname'] = self.params['runname'] classliststr = operations.intListToString(self.classlist) rctrunq['classnums'] = classliststr rctrunq['numiter'] = self.params['numiters'] rctrunq['maskrad'] = self.params['radius'] rctrunq['lowpassvol'] = self.params['lowpassvol'] rctrunq['highpasspart'] = self.params['highpasspart'] rctrunq['lowpasspart'] = self.params['lowpasspart'] rctrunq['median'] = self.params['median'] rctrunq['description'] = self.params['description'] rctrunq['path'] = appiondata.ApPathData(path=os.path.abspath(self.params['rundir'])) rctrunq['alignstack'] = self.alignstackdata rctrunq['tiltstack'] = apStack.getOnlyStackData(self.params['tiltstackid']) rctrunq['numpart'] = self.numpart rctrunq['fsc_resolution'] = fscresq rctrunq['rmeasure_resolution'] = rmeasureq if self.params['commit'] is True: rctrunq.insert() ### insert 3d volume density densq = appiondata.Ap3dDensityData() densq['rctrun'] = rctrunq densq['path'] = appiondata.ApPathData(path=os.path.dirname(os.path.abspath(volfile))) densq['name'] = os.path.basename(volfile) densq['hidden'] = False densq['norm'] = True densq['symmetry'] = appiondata.ApSymmetryData.direct_query(25) densq['pixelsize'] = apStack.getStackPixelSizeFromStackId(self.params['tiltstackid'])*self.params['tiltbin'] densq['boxsize'] = self.getBoxSize() densq['lowpass'] = self.params['lowpassvol'] densq['highpass'] = self.params['highpasspart'] densq['mask'] = self.params['radius'] #densq['iterid'] = self.params['numiters'] densq['description'] = self.params['description'] densq['resolution'] = self.fscresolution densq['rmeasure'] = self.rmeasureresolution densq['session'] = apStack.getSessionDataFromStackId(self.params['tiltstackid']) densq['md5sum'] = apFile.md5sumfile(volfile) if self.params['commit'] is True: densq.insert() return
def hierarchClusterProcess(numpart=None,
factorlist=range(1, 5),
corandata="coran/corandata",
rundir=".",
dataext=".spi"):
"""
inputs:
coran data
number of particles
factor list
output directory
output:
dendrogram doc file
factorkey
"""
#apFile.removeFile(rundir+"/dendrogramdoc"+dataext)
factorstr, factorkey = operations.intListToString(factorlist)
dendrogramfile = rundir + "/dendrogramdoc" + factorkey + dataext
if os.path.isfile(dendrogramfile):
apDisplay.printMsg(
"Dendrogram file already exists, skipping processing " +
dendrogramfile)
return dendrogramfile
apDisplay.printMsg("Creating dendrogram file: " + dendrogramfile)
### do hierarchical clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=False, log=True)
mySpider.toSpider(
"CL HC",
spyder.fileFilter(corandata) + "_IMC", # path to coran data
factorstr, # factor string
)
## weight for each factor
for fact in factorlist:
mySpider.toSpiderQuiet("1.0")
minclasssize = "%.4f" % (numpart * 0.0001 + 2.0)
mySpider.toSpider(
"5", #use Ward's method
"T",
minclasssize,
rundir + "/dendrogram.ps", #dendrogram image file
"Y",
spyder.fileFilter(dendrogramfile), #dendrogram doc file
)
mySpider.close()
if not os.path.isfile(dendrogramfile):
apDisplay.printError(
"SPIDER dendrogram creation (CL HC) failed, too many particles??")
apImage.convertPostscriptToPng("cluster/dendrogram.ps", "dendrogram.png")
return dendrogramfile
def hierarchClusterProcess(numpart=None, factorlist=range(1,5),
corandata="coran/corandata", rundir=".", dataext=".spi"):
"""
inputs:
coran data
number of particles
factor list
output directory
output:
dendrogram doc file
factorkey
"""
#apFile.removeFile(rundir+"/dendrogramdoc"+dataext)
factorstr, factorkey = operations.intListToString(factorlist)
dendrogramfile = rundir+"/dendrogramdoc"+factorkey+dataext
if os.path.isfile(dendrogramfile):
apDisplay.printMsg("Dendrogram file already exists, skipping processing "+dendrogramfile)
return dendrogramfile
apDisplay.printMsg("Creating dendrogram file: "+dendrogramfile)
### do hierarchical clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=False, log=True)
mySpider.toSpider(
"CL HC",
spyder.fileFilter(corandata)+"_IMC", # path to coran data
factorstr, # factor string
)
## weight for each factor
for fact in factorlist:
mySpider.toSpiderQuiet("1.0")
minclasssize = "%.4f" % (numpart*0.0001+2.0)
mySpider.toSpider(
"5", #use Ward's method
"T", minclasssize, rundir+"/dendrogram.ps", #dendrogram image file
"Y", spyder.fileFilter(dendrogramfile), #dendrogram doc file
)
mySpider.close()
if not os.path.isfile(dendrogramfile):
apDisplay.printError("SPIDER dendrogram creation (CL HC) failed, too many particles??")
apImage.convertPostscriptToPng("cluster/dendrogram.ps", "dendrogram.png")
return dendrogramfile
def insertRctRun(self, volfile):
### setup resolutions
fscresq = appiondata.ApResolutionData()
fscresq['type'] = "fsc"
fscresq['half'] = self.fscresolution
fscresq['fscfile'] = "fscdata" + self.timestamp + ".fsc"
rmeasureq = appiondata.ApResolutionData()
rmeasureq['type'] = "rmeasure"
rmeasureq['half'] = self.rmeasureresolution
rmeasureq['fscfile'] = None
### insert rct run data
rctrunq = appiondata.ApRctRunData()
rctrunq['runname'] = self.params['runname']
classliststr = operations.intListToString(self.classlist)
rctrunq['classnums'] = classliststr
rctrunq['numiter'] = self.params['numiters']
rctrunq['maskrad'] = self.params['radius']
rctrunq['lowpassvol'] = self.params['lowpassvol']
rctrunq['highpasspart'] = self.params['highpasspart']
rctrunq['lowpasspart'] = self.params['lowpasspart']
rctrunq['median'] = self.params['median']
rctrunq['description'] = self.params['description']
rctrunq['path'] = appiondata.ApPathData(
path=os.path.abspath(self.params['rundir']))
rctrunq['alignstack'] = self.alignstackdata
rctrunq['tiltstack'] = apStack.getOnlyStackData(
self.params['tiltstackid'])
rctrunq['numpart'] = self.numpart
rctrunq['fsc_resolution'] = fscresq
rctrunq['rmeasure_resolution'] = rmeasureq
if self.params['commit'] is True:
rctrunq.insert()
### insert 3d volume density
densq = appiondata.Ap3dDensityData()
densq['rctrun'] = rctrunq
densq['path'] = appiondata.ApPathData(
path=os.path.dirname(os.path.abspath(volfile)))
densq['name'] = os.path.basename(volfile)
densq['hidden'] = False
densq['norm'] = True
densq['symmetry'] = appiondata.ApSymmetryData.direct_query(25)
densq['pixelsize'] = apStack.getStackPixelSizeFromStackId(
self.params['tiltstackid']) * self.params['tiltbin']
densq['boxsize'] = self.getBoxSize()
densq['lowpass'] = self.params['lowpassvol']
densq['highpass'] = self.params['highpasspart']
densq['mask'] = self.params['radius']
#densq['iterid'] = self.params['numiters']
densq['description'] = self.params['description']
densq['resolution'] = self.fscresolution
densq['rmeasure'] = self.rmeasureresolution
densq['session'] = apStack.getSessionDataFromStackId(
self.params['tiltstackid'])
densq['md5sum'] = apFile.md5sumfile(volfile)
if self.params['commit'] is True:
densq.insert()
return
def kmeansCluster(alignedstack, numpart=None, numclasses=40, timestamp=None,
factorlist=range(1,5), corandata="coran/corandata", dataext=".spi"):
"""
inputs:
outputs:
"""
if timestamp is None:
timestamp = apParam.makeTimestamp()
if alignedstack[-4:] == dataext:
alignedstack = alignedstack[:-4]
rundir = "cluster"
classavg = rundir+"/"+("classavgstack_%s_%03d" % (timestamp, numclasses))
classvar = rundir+"/"+("classvarstack_%s_%03d" % (timestamp, numclasses))
apParam.createDirectory(rundir)
for i in range(numclasses):
apFile.removeFile(rundir+("/classdoc%04d" % (i+1))+dataext)
apFile.removeFile(rundir+("/allclassesdoc%04d" % (numclasses))+dataext)
### make list of factors
factorstr, factorkey = operations.intListToString(factorlist)
### do k-means clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=True, log=False)
mySpider.toSpider(
"CL KM",
corandata+"_IMC", # path to coran data
str(numclasses), # num classes
factorstr, # factor string
)
## weight for each factor
for fact in factorlist:
mySpider.toSpiderQuiet("1.0")
randnum = (int(random.random()*1000) + 1)
mySpider.toSpider(
str(randnum),
rundir+"/classdoc_"+timestamp+"_****", # class doc file
rundir+("/allclassesdoc%04d" % (numclasses)), #clusterdoc file
)
mySpider.close()
### delete existing files
sys.stderr.write("delete existing files")
for dext in (".hed", ".img", dataext):
apFile.removeFile(classavg+dext)
apFile.removeFile(classvar+dext)
print ""
mySpider = spyder.SpiderSession(dataext=dataext, logo=True, log=False)
### create class averages
apDisplay.printMsg("Averaging particles into classes")
for i in range(numclasses):
classnum = i+1
mySpider.toSpiderQuiet(
"AS R",
spyder.fileFilter(alignedstack)+"@******",
rundir+("/classdoc_"+timestamp+"_%04d" % (classnum)),
"A",
(classavg+"@%04d" % (classnum)),
(classvar+"@%04d" % (classnum)),
)
if classnum % 10 == 0:
sys.stderr.write(".")
time.sleep(1)
mySpider.close()
### convert to IMAGIC
emancmd = "proc2d "+classavg+".spi "+classavg+".hed"
apEMAN.executeEmanCmd(emancmd, verbose=False, showcmd=True)
emancmd = "proc2d "+classvar+".spi "+classvar+".hed"
apEMAN.executeEmanCmd(emancmd, verbose=False, showcmd=True)
return classavg,classvar
def start(self):
### get original aligned stack name
astack = self.analysisdata['alignstack']['imagicfile']
### spider has problems with file name if it includes an "x#"
astack = re.sub(r'x(\d)',r'x-\1',astack)
### get original align stack
imagicalignedstack = os.path.join(self.analysisdata['alignstack']['path']['path'],
astack)
alignedstack = re.sub("\.", "_", imagicalignedstack)+".spi"
while os.path.isfile(alignedstack):
apFile.removeFile(alignedstack)
emancmd = "proc2d %s %s spiderswap"%(imagicalignedstack, alignedstack)
apEMAN.executeEmanCmd(emancmd, showcmd=True, verbose=True)
### get database information
numpart = self.analysisdata['alignstack']['num_particles']
corandata = os.path.join(self.analysisdata['path']['path'],"coran/corandata")
### parse factor list
factorlist = self.params['factorstr'].split(",")
factorstr, factorkey = operations.intListToString(factorlist)
factorstr = re.sub(",", ", ", factorstr)
apDisplay.printMsg("using factorlist "+factorstr)
if len(factorlist) > self.analysisdata['coranrun']['num_factors']:
apDisplay.printError("Requested factor list is longer than available factors")
if self.params['commit'] is True:
self.insertClusterRun(insert=True)
else:
apDisplay.printWarning("not committing results to DB")
numclasslist = self.params['numclasslist'].split(",")
if self.params['method'] != "kmeans":
rundir = "cluster"
apParam.createDirectory(rundir)
### step 1: use coran data to create hierarchy
dendrogramfile = classification.hierarchClusterProcess(numpart, factorlist, corandata, rundir, dataext=".spi")
### step 2: asssign particles to groups based on hierarchy
for item in numclasslist:
t0 = time.time()
if not item or not re.match("^[0-9]+$", item):
continue
numclass = int(item)
apDisplay.printColor("\n============================\nprocessing class averages for "
+str(numclass)+" classes\n============================\n", "green")
#run the classification
if self.params['method'] == "kmeans":
apDisplay.printMsg("Using the k-means clustering method")
classavg,classvar = classification.kmeansCluster(alignedstack, numpart, numclasses=numclass,
timestamp=self.timestamp, factorlist=factorlist, corandata=corandata, dataext=".spi")
else:
apDisplay.printMsg("Using the hierarch clustering method")
classavg,classvar = classification.hierarchClusterClassify(alignedstack, dendrogramfile, numclass,
self.timestamp, rundir, dataext=".spi")
#classavg,classvar = classification.hierarchCluster(alignedstack, numpart, numclasses=numclass,
# timestamp=self.timestamp, factorlist=factorlist, corandata=corandata, dataext=".spi")
if self.params['commit'] is True:
self.insertClusterStack(classavg, classvar, numclass, insert=True)
else:
apDisplay.printWarning("not committing results to DB")
apDisplay.printMsg("Completed "+str(numclass)+" classes in "+apDisplay.timeString(time.time()-t0))
def kmeansCluster(alignedstack,
numpart=None,
numclasses=40,
timestamp=None,
factorlist=range(1, 5),
corandata="coran/corandata",
dataext=".spi"):
"""
inputs:
outputs:
"""
if timestamp is None:
timestamp = apParam.makeTimestamp()
if alignedstack[-4:] == dataext:
alignedstack = alignedstack[:-4]
rundir = "cluster"
classavg = rundir + "/" + ("classavgstack_%s_%03d" %
(timestamp, numclasses))
classvar = rundir + "/" + ("classvarstack_%s_%03d" %
(timestamp, numclasses))
apParam.createDirectory(rundir)
for i in range(numclasses):
apFile.removeFile(rundir + ("/classdoc%04d" % (i + 1)) + dataext)
apFile.removeFile(rundir + ("/allclassesdoc%04d" % (numclasses)) + dataext)
### make list of factors
factorstr, factorkey = operations.intListToString(factorlist)
### do k-means clustering
mySpider = spyder.SpiderSession(dataext=dataext, logo=True, log=False)
mySpider.toSpider(
"CL KM",
corandata + "_IMC", # path to coran data
str(numclasses), # num classes
factorstr, # factor string
)
## weight for each factor
for fact in factorlist:
mySpider.toSpiderQuiet("1.0")
randnum = (int(random.random() * 1000) + 1)
mySpider.toSpider(
str(randnum),
rundir + "/classdoc_" + timestamp + "_****", # class doc file
rundir + ("/allclassesdoc%04d" % (numclasses)), #clusterdoc file
)
mySpider.close()
### delete existing files
sys.stderr.write("delete existing files")
for dext in (".hed", ".img", dataext):
apFile.removeFile(classavg + dext)
apFile.removeFile(classvar + dext)
print ""
mySpider = spyder.SpiderSession(dataext=dataext, logo=True, log=False)
### create class averages
apDisplay.printMsg("Averaging particles into classes")
for i in range(numclasses):
classnum = i + 1
mySpider.toSpiderQuiet(
"AS R",
spyder.fileFilter(alignedstack) + "@******",
rundir + ("/classdoc_" + timestamp + "_%04d" % (classnum)),
"A",
(classavg + "@%04d" % (classnum)),
(classvar + "@%04d" % (classnum)),
)
if classnum % 10 == 0:
sys.stderr.write(".")
time.sleep(1)
mySpider.close()
### convert to IMAGIC
emancmd = "proc2d " + classavg + ".spi " + classavg + ".hed"
apEMAN.executeEmanCmd(emancmd, verbose=False, showcmd=True)
emancmd = "proc2d " + classvar + ".spi " + classvar + ".hed"
apEMAN.executeEmanCmd(emancmd, verbose=False, showcmd=True)
return classavg, classvar
def start(self):
### get original aligned stack name
astack = self.analysisdata['alignstack']['imagicfile']
### spider has problems with file name if it includes an "x#"
astack = re.sub(r'x(\d)', r'x-\1', astack)
### get original align stack
imagicalignedstack = os.path.join(
self.analysisdata['alignstack']['path']['path'], astack)
alignedstack = re.sub("\.", "_", imagicalignedstack) + ".spi"
while os.path.isfile(alignedstack):
apFile.removeFile(alignedstack)
emancmd = "proc2d %s %s spiderswap" % (imagicalignedstack,
alignedstack)
apEMAN.executeEmanCmd(emancmd, showcmd=True, verbose=True)
### get database information
numpart = self.analysisdata['alignstack']['num_particles']
corandata = os.path.join(self.analysisdata['path']['path'],
"coran/corandata")
### parse factor list
factorlist = self.params['factorstr'].split(",")
factorstr, factorkey = operations.intListToString(factorlist)
factorstr = re.sub(",", ", ", factorstr)
apDisplay.printMsg("using factorlist " + factorstr)
if len(factorlist) > self.analysisdata['coranrun']['num_factors']:
apDisplay.printError(
"Requested factor list is longer than available factors")
if self.params['commit'] is True:
self.insertClusterRun(insert=True)
else:
apDisplay.printWarning("not committing results to DB")
numclasslist = self.params['numclasslist'].split(",")
if self.params['method'] != "kmeans":
rundir = "cluster"
apParam.createDirectory(rundir)
### step 1: use coran data to create hierarchy
dendrogramfile = classification.hierarchClusterProcess(
numpart, factorlist, corandata, rundir, dataext=".spi")
### step 2: asssign particles to groups based on hierarchy
for item in numclasslist:
t0 = time.time()
if not item or not re.match("^[0-9]+$", item):
continue
numclass = int(item)
apDisplay.printColor(
"\n============================\nprocessing class averages for "
+ str(numclass) + " classes\n============================\n",
"green")
#run the classification
if self.params['method'] == "kmeans":
apDisplay.printMsg("Using the k-means clustering method")
classavg, classvar = classification.kmeansCluster(
alignedstack,
numpart,
numclasses=numclass,
timestamp=self.timestamp,
factorlist=factorlist,
corandata=corandata,
dataext=".spi")
else:
apDisplay.printMsg("Using the hierarch clustering method")
classavg, classvar = classification.hierarchClusterClassify(
alignedstack,
dendrogramfile,
numclass,
self.timestamp,
rundir,
dataext=".spi")
#classavg,classvar = classification.hierarchCluster(alignedstack, numpart, numclasses=numclass,
# timestamp=self.timestamp, factorlist=factorlist, corandata=corandata, dataext=".spi")
if self.params['commit'] is True:
self.insertClusterStack(classavg,
classvar,
numclass,
insert=True)
else:
apDisplay.printWarning("not committing results to DB")
apDisplay.printMsg("Completed " + str(numclass) + " classes in " +
apDisplay.timeString(time.time() - t0))
