def main(args):
usage = """python %s <intermediate phenotye>
Run a meta analysis of the top-ranked pairs for the different cohorts.
Save results in %s/meta_<intermediate phenotype>.sortedpvals.
""" % (args[0], rDir)
if len(args) == 2:
phenotypeName = args[1]
else:
print usage
sys.exit(0)
# Get the list of pairs to run
threshold = 1e-12
sigPairsList = set([])
for cohort in ['Dutch_MA', 'Dutch_MO', 'Finnish_MA', 'German_MA']:
with open("%s/%s_%s.sortedpvals" % (rDir, cohort, phenotypeName)) as f:
f.readline()
for line in f:
ls = line.split()
if float(ls[-1]) > threshold:
break
pair = [
'%s_%s_%s' % (ls[0], ls[1], ls[2]),
'%s_%s_%s' % (ls[3], ls[4], ls[5])
]
pair.sort()
pair = "/".join(pair)
sigPairsList.add(pair)
f.close()
sigPairsList = list(sigPairsList)
# Only keep pairs of SNPs that appear in all studies
for cohort in ['Dutch_MA', 'Dutch_MO', 'Finnish_MA', 'German_MA']:
f = open('%s/%s/%s_clean_%s.bim' %
(dataDirRoot, cohort, cohort, phenotypeName))
if not allSnps:
allSnps = set([line.split()[1] for line in f])
else:
allSnps = allSnps.intersection(set([line.split()[1]
for line in f]))
f.close()
print len(allSnps), "SNPs appear in all studies"
sigPairsList = [x for x in sigPairsList if \
((x.split("/")[0].split("_")[0] in allSnps) \
and (x.split("/")[1].split("_")[0] in allSnps))]
print len(sigPairsList), "pairs to run."
# Test pairs for each study
sigPairsDict = {pair: [] for pair in sigPairsList}
for cohort in ['Dutch_MA', 'Dutch_MO', 'Finnish_MA', 'German_MA']:
print "Testing for %s" % cohort
dataDir = '%s/%s' % (dataDirRoot, cohort)
phenoF = '%s/%s_clean_%s.phenoGlide' % (dataDir, cohort, phenotypeName)
bimF = '%s/%s_clean_%s.bim' % (dataDir, cohort, phenotypeName)
h5fname = '%s/%s_clean_%s.h5' % (dataDir, cohort, phenotypeName)
pheno = np.loadtxt(phenoF)
with open(bimF) as f:
snpsDict = {('%s_%s_%s' % \
(line.split()[1], line.split()[0], line.split()[3])):idx \
for (idx, line) in enumerate(f)}
f.close()
with tables.openFile(h5fname) as h5f:
for pair in sigPairsList:
pairSplit = pair.split("/")
snp1idx = snpsDict[pairSplit[0]]
snp2idx = snpsDict[pairSplit[1]]
snp1x = np.ma.masked_values(h5f.root.genotype[snp1idx], 3)
snp2x = np.ma.masked_values(h5f.root.genotype[snp2idx], 3)
sigPairsDict[pair].append(testPair(snp1x, snp2x, pheno))
h5f.close()
# Combine into a meta-analysis
pvalsDict = {}
for pair, outputs in sigPairsDict.iteritems():
betas = np.array([x[0] for x in outputs])
sesqs = np.array([x[1] for x in outputs])
betas[np.where(betas == 0)] = 1e-10
sesqs[np.where(sesqs <= 0)] = 1e-10
sesqinvs = 1. / sesqs
betam = np.sum(betas * sesqinvs) / np.sum(sesqinvs)
sesqm = np.sqrt(1. / np.sum(sesqinvs))
zmeta = betam / sesqm
pval = 2 * st.norm.sf(zmeta) # two-sided
try:
pvalsDict[pval].append([pair, zmeta])
except KeyError:
pvalsDict[pval] = [[pair, zmeta]]
# sort p-values and save
pvals = pvalsDict.keys()
pvals.sort()
with open('%s/meta_logistic_%s.sortedpvals' % (rDir, phenotypeName),
'w') as f:
f.write(
"SNP1 & chr & pos & SNP2 & chr & pos & meta z-score & meta pval \n"
)
for pval in pvals:
for [pair, zmeta] in pvalsDict[pval]:
[snp1idx, snp2idx] = pair.split("/")
snp1 = snp1idx
snp2 = snp2idx
f.write("%s & %s & %.2e & %.2e\n" % \
(" & ".join(snp1.split("_")),
" & ".join(snp2.split("_")),
zmeta, pval))
f.close()
def main(args):
usage = """python %s <glideIn> <snps list> <pheno> <confounders> <significant pairs of SNPs> <reevaluated pairs of SNPs>
Rerun linear regression evaluation on the significant pairs of SNPs.
E.g.: py rerun_confounder.py mydata_final_clean.glideIn mydata_final_clean.snpNames mydata_samples.sex mydata_final_clean.pheno nydata_clean.sortedpvals mydata_final_clean.sortedpvals.rerun_sex\n""" % args[
0]
if len(args) != 7:
sys.stderr.write(usage)
sys.exit(0)
glideInF = args[1]
snpsLstF = args[2]
cfdrsF = args[3]
phenoF = args[4]
sigPairF = args[5]
outputF = args[6]
threshold = 1e-11
pheno = np.loadtxt(phenoF)
confounders = np.loadtxt(cfdrsF)
f = open(snpsLstF)
snpsDict = {line.split()[0]: idx for (idx, line) in enumerate(f)}
f.close()
newPvalsDict = {} # pvalue, line to write
with open(sigPairF) as f:
f.readline()
for i, line in enumerate(f):
ls = line.split()
try:
snp1idx = snpsDict[ls[0]]
try:
snp2idx = snpsDict[ls[1]]
snp1x = readSnp(snp1idx, glideInF)
snp2x = readSnp(snp2idx, glideInF)
pval, pvalStr = testPair(snp1x, snp2x, pheno, confounders)
line_to_write = "%s\t%s\n" % ("\t".join(ls), pvalStr)
if not newPvalsDict.has_key(pval):
newPvalsDict[pval] = [line_to_write]
else:
newPvalsDict[pval].append(line_to_write)
except KeyError:
print "Didn't find %s in SNPs list" % ls[3]
sys.exit(-1)
except KeyError:
print "Didn't find %s in SNPs list" % ls[0]
sys.exit(-1)
f.close()
with open(outputF, 'w') as g:
g.write("SNP1\tchr1\tpos1\tSNP2\tchr2\tpos2\t")
g.write(
"t-testGLIDE\tpvalGLIDE(intercept)\tpvalGLIDE(x1)\tpvalGLIDE(x2)\tpvalGLIDE(x1:x2)"
)
g.write(
"\tpval(intercept)\tpval(x1)\tpval(x2)\tpval(confounder(s)\tpval(x1:x2)\n"
)
sortedPvals = newPvalsDict.keys()
sortedPvals.sort()
for pval in sortedPvals:
for line_to_write in newPvalsDict[pval]:
g.write(line_to_write)
g.close()
def main(args):
usage = """python %s <glideIn> <snps list> <pheno> <confounders> <significant pairs of SNPs> <reevaluated pairs of SNPs>
Rerun linear regression evaluation on the significant pairs of SNPs.
E.g.: py rerun_confounder.py mydata_final_clean.h5 mydata_final_clean.bim mydata_samples.sex mydata_final_clean.phenoGlide mydata_final_clean.sortedpvals mydata_final_clean.sortedpvals.rerun_sex\n""" % args[
0]
if len(args) != 7:
sys.stderr.write(usage)
sys.exit(0)
h5fname = args[1]
bimF = args[2]
cfdrsF = args[3]
phenoF = args[4]
sigPairF = args[5]
outputF = args[6]
threshold = 1e-11
pheno = np.loadtxt(phenoF)
confounders = np.loadtxt(cfdrsF)
with open(bimF) as f:
snpsDict = {('%s\t%s\t%s' % \
(line.split()[1], line.split()[0], line.split()[3])):idx \
for (idx, line) in enumerate(f)}
newPvalsDict = {} # pvalue, line to write
with tables.openFile(h5fname) as h5f, open(sigPairF) as f:
print "%d SNPs x %d samples" % (h5f.root.genotype.shape[0],
h5f.root.genotype.shape[1])
f.readline()
for i, line in enumerate(f):
ls = line.split()
try:
snp1idx = snpsDict['%s\t%s\t%s' % (ls[0], ls[1], ls[2])]
try:
snp2idx = snpsDict['%s\t%s\t%s' % (ls[3], ls[4], ls[5])]
snp1x = np.ma.masked_values(h5f.root.genotype[snp1idx], 3)
snp2x = np.ma.masked_values(h5f.root.genotype[snp2idx], 3)
pval, pvalStr = testPair(snp1x, snp2x, pheno, confounders)
line_to_write = "%s\t%s\n" % ("\t".join(ls), pvalStr)
if not newPvalsDict.has_key(pval):
newPvalsDict[pval] = [line_to_write]
else:
newPvalsDict[pval].append(line_to_write)
except KeyError:
print "Didn't find %s in .bim file" % ls[3]
sys.exit(-1)
except KeyError:
print "Didn't find %s in .bim file" % ls[0]
sys.exit(-1)
f.close()
h5f.close()
with open(outputF, 'w') as g:
g.write("SNP1\tchr1\tpos1\tSNP2\tchr2\tpos2\t")
g.write(
"t-testGLIDE\tpvalGLIDE(intercept)\tpvalGLIDE(x1)\tpvalGLIDE(x2)\tpvalGLIDE(x1:x2)"
)
g.write(
"\tpval(intercept)\tpval(x1)\tpval(x2)\tpval(confounder(s))\tpval(x1:x2)\n"
)
sortedPvals = newPvalsDict.keys()
sortedPvals.sort()
for pval in sortedPvals:
for line_to_write in newPvalsDict[pval]:
g.write(line_to_write)
g.close()