def analyzeColumn(
summaryDataColumn, listPdbs, outputFileName, numTests=1000000):
'''takes a column of summary data, computes average and permutes, outputs'''
origDiffMean = getDiffMean(summaryDataColumn, listPdbs)
pValCounts = [0., 0.] # above, below
for test in range(numTests):
newLists = statistics.permuteLists(listPdbs)
testMean = getDiffMean(summaryDataColumn, newLists)
if testMean >= origDiffMean:
pValCounts[0] += 1.
if testMean <= origDiffMean:
pValCounts[1] += 1.
pVals = [pValCount/float(numTests) for pValCount in pValCounts]
outputFile = open(outputFileName, 'w')
outputFile.write("origDiffMean\tmean1\tmean2\tpValAbove\tpValBelow\n")
outputFile.write(str(origDiffMean) + "\t")
outputFile.write(str(getMean(summaryDataColumn, listPdbs[0])) + "\t")
outputFile.write(str(getMean(summaryDataColumn, listPdbs[1])) + "\t")
for pVal in pVals:
outputFile.write(str(pVal) + "\t")
outputFile.write("\n")
outputFile.close()
def makeCompareResidueReportAlternate(
pdbs, outputFilename="residue.bfactor", numTests=9999,
correctionAll=0., correctionBeta=0.):
'''different way to do p-vals, instead of permuting all data, permute the
pairs of hyp/meso pdb files.'''
residueNames = aminoAcid3Codes # for now ignore what is in the files
fileTemp2 = open(outputFilename + ".pvals.txt", 'w')
fileTemp2.write("ResidueName DiffMeans MeanA MeanB pValAbove pValBelow\n")
fileTemp3 = open(outputFilename + ".pvals.beta.txt", 'w')
fileTemp3.write("ResidueName DiffMeans MeanA MeanB pValAbove pValBelow\n")
#first find means
means = [{}, {}]
betaMeans = [{}, {}]
overallList = [[], []]
overallBetaList = [[], []]
totalMeans, totalBetaMeans = [0., 0.], [0., 0.]
for code in residueNames:
betaKsLists = [[], []]
for lindex in range(2): # either a or b
totalList, betaList = [], []
for pdbResidues in pdbs[lindex]:
if code in pdbResidues:
for atomValues in pdbResidues[code].values():
totalList.extend(atomValues)
means[lindex][code] = statistics.computeMean(totalList)
for pdbResidues in pdbs[lindex]:
if code in pdbResidues:
betaList.extend(pdbResidues[code][carbonBetaCodes[code]])
betaKsLists[lindex] = betaList
betaMeans[lindex][code] = statistics.computeMean(betaList)
overallList[lindex].extend(totalList)
overallBetaList[lindex].extend(betaList)
#use betaKsLists to compute ks stuff
for lindex in range(2): # either a or b
totalMeans[lindex] = statistics.computeMean(overallList[lindex])
totalBetaMeans[lindex] = statistics.computeMean(overallBetaList[lindex])
#print means, betaMeans
pValueCounts = [{}, {}] # first is above, second is below
pValueBetaCounts = [{}, {}]
for code in residueNames+["ALL"]: # initialize counts, even for overall total
for aboveBelow in range(2):
pValueCounts[aboveBelow][code] = 1
pValueBetaCounts[aboveBelow][code] = 1
for test in xrange(numTests):
testMeans = [{}, {}]
testBetaMeans = [{}, {}]
overallList = [[], []]
overallBetaList = [[], []]
totalTestMeans, totalTestBetaMeans = [0., 0.], [0., 0.]
newPdbs = statistics.permuteLists(pdbs)
for code in residueNames:
for lindex in range(2): # either a or b
totalList, betaList = [], []
for pdbResidues in newPdbs[lindex]:
if code in pdbResidues:
for atomValues in pdbResidues[code].values():
totalList.extend(atomValues)
testMeans[lindex][code] = statistics.computeMean(totalList)
for pdbResidues in newPdbs[lindex]:
if code in pdbResidues:
betaList.extend(pdbResidues[code][carbonBetaCodes[code]])
testBetaMeans[lindex][code] = statistics.computeMean(betaList)
overallList[lindex].extend(totalList)
overallBetaList[lindex].extend(betaList)
for lindex in range(2): # either a or b
totalTestMeans[lindex] = statistics.computeMean(overallList[lindex])
totalTestBetaMeans[lindex] = \
statistics.computeMean(overallBetaList[lindex])
for code in residueNames: # calc pval for each residue
testMeanDiff = testMeans[0][code] - testMeans[1][code]
origMeanDiff = means[0][code] - means[1][code] - correctionAll
if origMeanDiff <= testMeanDiff:
pValueCounts[0][code] += 1
if origMeanDiff >= testMeanDiff:
pValueCounts[1][code] += 1
testMeanDiff = testBetaMeans[0][code] - testBetaMeans[1][code]
origMeanDiff = betaMeans[0][code] - betaMeans[1][code] - correctionBeta
if origMeanDiff <= testMeanDiff:
pValueBetaCounts[0][code] += 1
if origMeanDiff >= testMeanDiff:
pValueBetaCounts[1][code] += 1
code = "ALL" # fake residue name for overall
testMeanDiff = totalTestMeans[0] - totalTestMeans[1]
origMeanDiff = totalMeans[0] - totalMeans[1] - correctionAll
if origMeanDiff <= testMeanDiff:
pValueCounts[0][code] += 1
if origMeanDiff >= testMeanDiff:
pValueCounts[1][code] += 1
testMeanDiff = totalTestBetaMeans[0] - totalTestBetaMeans[1] - \
correctionBeta
origMeanDiff = totalBetaMeans[0] - totalBetaMeans[1]
if origMeanDiff <= testMeanDiff:
pValueBetaCounts[0][code] += 1
if origMeanDiff >= testMeanDiff:
pValueBetaCounts[1][code] += 1
for code in residueNames: # output time
fileTemp2.write(code + " " + str(means[0][code]-means[1][code]) + " ")
fileTemp2.write(str(means[0][code]) + " " + str(means[1][code]) + " ")
fileTemp2.write(str(pValueCounts[0][code]/float(1+numTests)) + " ")
fileTemp2.write(str(pValueCounts[1][code]/float(1+numTests)) + " ")
fileTemp2.write("\n")
fileTemp3.write(
code + " " + str(betaMeans[0][code]-betaMeans[1][code]) + " ")
fileTemp3.write(
str(betaMeans[0][code]) + " " + str(betaMeans[1][code]) + " ")
fileTemp3.write(str(pValueBetaCounts[0][code]/float(1+numTests)) + " ")
fileTemp3.write(str(pValueBetaCounts[1][code]/float(1+numTests)) + " ")
fileTemp3.write("\n")
code = "ALL" # fake for overall
fileTemp2.write("ALL " + str(totalMeans[0]-totalMeans[1]) + " ")
fileTemp2.write(str(totalMeans[0]) + " " + str(totalMeans[1]) + " ")
fileTemp2.write(str(pValueCounts[0][code]/float(1+numTests)) + " ")
fileTemp2.write(str(pValueCounts[1][code]/float(1+numTests)) + " ")
fileTemp2.write("\n")
fileTemp3.write("ALL " + str(totalBetaMeans[0]-totalBetaMeans[1]) + " ")
fileTemp3.write(str(totalBetaMeans[0]) + " " + str(totalBetaMeans[1]) + " ")
fileTemp3.write(str(pValueBetaCounts[0][code]/float(1+numTests)) + " ")
fileTemp3.write(str(pValueBetaCounts[1][code]/float(1+numTests)) + " ")
fileTemp3.write("\n")
fileTemp2.close()
fileTemp3.close()
return totalMeans[0]-totalMeans[1], totalBetaMeans[0]-totalBetaMeans[1]
