def hail_calculation(ds):
rrm = hl.realized_relationship_matrix(ds['GT'])
fn = utils.new_temp_file(suffix='.tsv')
rrm.export_tsv(fn)
data = []
with open(utils.uri_path(fn)) as f:
f.readline()
for line in f:
row = line.strip().split()
data.append(list(map(float, row)))
return np.array(data)
def test_linear_mixed_model_function(self):
n, f, m = 4, 2, 3
y = np.array([0.0, 1.0, 8.0, 9.0])
x = np.array([[1.0, 0.0],
[1.0, 2.0],
[1.0, 1.0],
[1.0, 4.0]])
z = np.array([[0.0, 0.0, 1.0],
[0.0, 1.0, 2.0],
[1.0, 2.0, 0.0],
[2.0, 0.0, 1.0]])
p_path = utils.new_temp_file()
def make_call(gt):
if gt == 0.0:
return hl.Call([0, 0])
if gt == 1.0:
return hl.Call([0, 1])
if gt == 2.0:
return hl.Call([1, 1])
data = [{'v': j, 's': i, 'y': y[i], 'x1': x[i, 1], 'zt': make_call(z[i, j])}
for i in range(n) for j in range(m)]
ht = hl.Table.parallelize(data, hl.dtype('struct{v: int32, s: int32, y: float64, x1: float64, zt: tcall}'))
mt = ht.to_matrix_table(row_key=['v'], col_key=['s'], col_fields=['x1', 'y'])
colsort = np.argsort(mt.key_cols_by().s.collect()).tolist()
mt = mt.choose_cols(colsort)
rrm = hl.realized_relationship_matrix(mt.zt).to_numpy()
# kinship path agrees with from_kinship
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1], k=rrm, p_path=p_path, overwrite=True)
model0, p0 = LinearMixedModel.from_kinship(y, x, rrm, p_path, overwrite=True)
assert model0._same(model)
assert np.allclose(p0, p)
# random effects path with standardize=True agrees with low-rank rrm
s0, u0 = np.linalg.eigh(rrm)
s0 = np.flip(s0, axis=0)[:m]
p0 = np.fliplr(u0).T[:m, :]
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1], z_t=mt.zt.n_alt_alleles(), p_path=p_path, overwrite=True)
model0 = LinearMixedModel(p0 @ y, p0 @ x, s0, y, x, p_path=p_path)
assert model0._same(model)
# random effects path with standardize=False agrees with from_random_effects
model0, p0 = LinearMixedModel.from_random_effects(y, x, z, p_path, overwrite=True)
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1], z_t=mt.zt.n_alt_alleles(), p_path=p_path, overwrite=True, standardize=False)
assert model0._same(model)
assert np.allclose(p0, p.to_numpy())
def test_linear_mixed_model_function(self):
n, f, m = 4, 2, 3
y = np.array([0.0, 1.0, 8.0, 9.0])
x = np.array([[1.0, 0.0], [1.0, 2.0], [1.0, 1.0], [1.0, 4.0]])
z = np.array([[0.0, 0.0, 1.0], [0.0, 1.0, 2.0], [1.0, 2.0, 0.0],
[2.0, 0.0, 1.0]])
p_path = utils.new_temp_file()
def make_call(gt):
if gt == 0.0:
return hl.Call([0, 0])
if gt == 1.0:
return hl.Call([0, 1])
if gt == 2.0:
return hl.Call([1, 1])
data = [{
'v': j,
's': i,
'y': y[i],
'x1': x[i, 1],
'zt': make_call(z[i, j])
} for i in range(n) for j in range(m)]
ht = hl.Table.parallelize(
data,
hl.dtype(
'struct{v: int32, s: int32, y: float64, x1: float64, zt: tcall}'
))
mt = ht.to_matrix_table(row_key=['v'],
col_key=['s'],
col_fields=['x1', 'y'])
colsort = np.argsort(mt.key_cols_by().s.collect()).tolist()
mt = mt.choose_cols(colsort)
rrm = hl.realized_relationship_matrix(mt.zt).to_numpy()
# kinship path agrees with from_kinship
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1],
k=rrm,
p_path=p_path,
overwrite=True)
model0, p0 = LinearMixedModel.from_kinship(y,
x,
rrm,
p_path,
overwrite=True)
assert model0._same(model)
assert np.allclose(p0, p)
# random effects path with standardize=True agrees with low-rank rrm
s0, u0 = np.linalg.eigh(rrm)
s0 = np.flip(s0, axis=0)[:m]
p0 = np.fliplr(u0).T[:m, :]
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1],
z_t=mt.zt.n_alt_alleles(),
p_path=p_path,
overwrite=True)
model0 = LinearMixedModel(p0 @ y, p0 @ x, s0, y, x, p_path=p_path)
assert model0._same(model)
# random effects path with standardize=False agrees with from_random_effects
model0, p0 = LinearMixedModel.from_random_effects(y,
x,
z,
p_path,
overwrite=True)
model, p = hl.linear_mixed_model(mt.y, [1, mt.x1],
z_t=mt.zt.n_alt_alleles(),
p_path=p_path,
overwrite=True,
standardize=False)
assert model0._same(model)
assert np.allclose(p0, p.to_numpy())