def scan(bfile,
Y,
cov,
null,
wnds,
minSnps,
i0,
i1,
perm_i,
resfile,
F,
colCovarType_r='lowrank',
rank_r=1,
factr=1e7):
if perm_i is not None:
print(('Generating permutation (permutation %d)' % perm_i))
np.random.seed(perm_i)
perm = np.random.permutation(Y.shape[0])
mtSet = limix.MTSet(Y=Y,
S_R=cov['eval'],
U_R=cov['evec'],
F=F,
rank=rank_r)
mtSet.setNull(null)
bim = plink_reader.readBIM(bfile, usecols=(0, 1, 2, 3))
fam = plink_reader.readFAM(bfile, usecols=(0, 1))
print('fitting model')
wnd_file = csv.writer(open(resfile, 'w'), delimiter='\t')
for wnd_i in range(i0, i1):
print(
('.. window %d - (%d, %d-%d) - %d snps' %
(wnd_i, int(wnds[wnd_i, 1]), int(
wnds[wnd_i, 2]), int(wnds[wnd_i, 3]), int(wnds[wnd_i, -1]))))
if int(wnds[wnd_i, -1]) < minSnps:
print('SKIPPED: number of snps lower than minSnps')
continue
#RV = bed.read(PositionRange(int(wnds[wnd_i,-2]),int(wnds[wnd_i,-1])))
RV = plink_reader.readBED(bfile,
useMAFencoding=True,
blocksize=1,
start=int(wnds[wnd_i, 4]),
nSNPs=int(wnds[wnd_i, 5]),
order='F',
standardizeSNPs=False,
ipos=2,
bim=bim,
fam=fam)
Xr = RV['snps']
if perm_i is not None:
Xr = Xr[perm, :]
Xr = np.ascontiguousarray(Xr)
rv = mtSet.optimize(Xr, factr=factr)
line = np.concatenate([wnds[wnd_i, :], rv['LLR']])
wnd_file.writerow(line)
pass
def test_mtSetPC_fixed(self):
fbasename = 'mtSetPC_fixed'
setTest = limix.MTSet(Y=self.Y, F=self.Xr[:, :2])
optInfo = setTest.optimize(self.Xr)
ext = {'Cr': optInfo['Cr'], 'Cn': optInfo['Cn']}
if self.write: self.saveStuff(fbasename, ext)
RV = self.assess(fbasename, ext)
self.assertTrue(RV)
def test_mtSetPCnull_fixed(self):
fbasename = 'mtSetPCnull_fixed'
setTest = limix.MTSet(Y=self.Y, F=self.Xr)
nullMTInfo = setTest.fitNull(cache=False)
ext = {'Cg': nullMTInfo['Cg'], 'Cn': nullMTInfo['Cn']}
if self.write: self.saveStuff(fbasename, ext)
RV = self.assess(fbasename, ext)
self.assertTrue(RV)
def test_mtSet_eigenCache(self):
fbasename = 'mtSet_base'
S, U = LA.eigh(self.XX)
setTest = limix.MTSet(Y=self.Y, S_R=S, U_R=U)
optInfo = setTest.optimize(self.Xr)
ext = {'Cr': optInfo['Cr'], 'Cg': optInfo['Cg'], 'Cn': optInfo['Cn']}
RV = self.assess(fbasename, ext)
self.assertTrue(RV)
def test_mtSet_base(self):
fbasename = 'mtSet_base'
setTest = limix.MTSet(self.Y, R=self.XX)
optInfo = setTest.optimize(self.Xr)
ext = {'Cr': optInfo['Cr'], 'Cg': optInfo['Cg'], 'Cn': optInfo['Cn']}
if self.write: self.saveStuff(fbasename, ext)
RV = self.assess(fbasename, ext)
self.assertTrue(RV)
def test_mtSetNull_eigenCache(self):
fbasename = 'mtSetNull'
S, U = LA.eigh(self.XX)
setTest = limix.MTSet(Y=self.Y, S_R=S, U_R=U)
nullMTInfo = setTest.fitNull(cache=False)
ext = {'Cg': nullMTInfo['Cg'], 'Cn': nullMTInfo['Cn']}
if self.write: self.saveStuff(fbasename, ext)
RV = self.assess(fbasename, ext)
self.assertTrue(RV)
def fit_null(Y, S_XX, U_XX, nfile, F):
"""
fit null model
Y NxP phenotype matrix
S_XX eigenvalues of the relatedness matrix
U_XX eigen vectors of the relatedness matrix
"""
mtSet = limix.MTSet(Y, S_R=S_XX, U_R=U_XX, F=F)
RV = mtSet.fitNull(cache=False)
params = np.array([RV['params0_g'], RV['params0_n']])
np.savetxt(nfile + '.p0', params)
np.savetxt(nfile + '.nll0', RV['NLL0'])
np.savetxt(nfile + '.cg0', RV['Cg'])
np.savetxt(nfile + '.cn0', RV['Cn'])
if __name__ == '__main__':
# generate data
N = 5000
f = 10
P = 4
K = 3
Y = sp.randn(N, P)
G = 1. * (sp.rand(N, f) < 0.2)
X = 1. * (sp.rand(N, f) < 0.2)
R = covar_rescale(sp.dot(X, X.T))
R += 1e-4 * sp.eye(N)
S_R, U_R = la.eigh(R)
F = sp.rand(N, K)
# ipdb.set_trace()
# mtSet not PC
if 1:
# mtSetPC
setTest = limix.MTSet(Y=Y, F=F)
else:
# mtSet
setTest = limix.MTSet(Y=Y, S_R=S_R, U_R=U_R)
nullMTInfo = setTest.fitNull(cache=False)
nullSTInfo = setTest.fitNullTraitByTrait(cache=False)
optInfo = setTest.optimize(G)
optInfo = setTest.optimizeTraitByTrait(G)
def scan(bfile,
Y,
cov,
null,
sets,
i0,
i1,
perm_i,
resfile,
F,
colCovarType_r='lowrank',
rank_r=1,
factr=1e7,
unique_variants=False,
standardize=False):
if perm_i is not None:
print(('Generating permutation (permutation %d)' % perm_i))
np.random.seed(perm_i)
perm = np.random.permutation(Y.shape[0])
mtSet = limix.MTSet(Y=Y,
S_R=cov['eval'],
U_R=cov['evec'],
F=F,
rank=rank_r)
mtSet.setNull(null)
reader = BedReader(bfile)
wnd_ids = sp.arange(i0, i1)
LLR = sp.zeros(sets.shape[0])
for wnd_i in wnd_ids:
_set = sets.ix[wnd_i]
print('.. set %d: %s' % (wnd_i, _set['setid']))
Xr = reader.getGenotypes(pos_start=_set['start'],
pos_end=_set['end'],
chrom=_set['chrom'],
impute=True)
if unique_variants:
Xr = f_uni_variants(Xr)
if standardize:
Xr -= Xr.mean(0)
Xr /= Xr.std(0)
else:
# encoding minor as 0
p = 0.5 * Xr.mean(0)
Xr[:, p > 0.5] = 2 - Xr[:, p > 0.5]
if perm_i is not None:
Xr = Xr[perm, :]
# multi trait set test fit
RV = mtSet.optimize(Xr, factr=factr)
LLR[wnd_i] = RV['LLR'][0]
# export results
sets['LLR'] = LLR
sets.to_csv(resfile, sep='\t', index=False)
