def __init__(
self,
entries_path,
phenotype_path=None,
covariate_path=None,
ibd_path=None,
one_hot_covariates=[],
bool_covariates=[],
drop_covariates=[],
):
"""
Creates a Population class instance.
:param pedigree: A Pedigree object
:param pheotypes: Series of phenotypes indexed by IID
:param covariates: Dataframe of covariates indexed by IID
:param covariances: List of covariance matrices aligned to Pedigree object
entries list
"""
self.entries = self._load_entries(entries_path)
self.entry_map = pd.Series(self.entries.index,
index=self.entries.values,
name='id')
self.phenotype = self._load_phenotype(phenotype_path)
self.covariate_info = None
self.one_hot_covariates = one_hot_covariates
self.drop_covariates = drop_covariates
self.bool_covariates = bool_covariates
self.covariates = self._load_covariates(covariate_path)
self.ibd = load_sparse_csr(ibd_path) if ibd_path else None
self.he = None
self.results = None
self.informative_indices = self._informative_indices()
def _load_ibd(self, ibd_path):
"""
Loads in IBD matrix
:param ibd_path: Path to IBD file
"""
if ibd_path is None:
return None
ibd = load_sparse_csr(ibd_path)
return ibd[self.entries.to_numpy()][:, self.entries.to_numpy()]
def SciLMM(
simulate=False,
sample_size=100000,
sparsity_factor=0.001,
gen_exp=1.4,
init_keep_rate=0.8,
fam=None,
ibd=False,
epis=False,
dom=False,
ibd_path=False,
epis_path=False,
dom_path=False,
gen_y=False,
y=None,
cov=None,
he=False,
lmm=False,
reml=False,
sim_num=100,
intercept=False,
verbose=False,
output_folder=".",
remove_cycles=False,
check_num_parents=False,
):
if ibd or epis or dom:
if not os.path.exists(output_folder):
raise Exception("The output folder does not exists")
if he or lmm:
if y is None and gen_y is False:
raise Exception("Can't estimate without a target value (--y)")
rel, interest_in_relevant = None, None
if fam:
rel_org, sex, interest, entries_dict = read_fam(fam_file_path=fam)
rel, interest_in_relevant = organize_rel(
rel_org,
interest,
remove_cycles=remove_cycles,
check_num_parents=check_num_parents,
)
# TODO: have to do sex as well in this version
entries_list = np.array(list(entries_dict.values()))[
interest_in_relevant
]
np.save(os.path.join(output_folder, "entries_ids.npy"), entries_list)
elif simulate:
if sample_size <= 0:
raise Exception("Sample size should be a positive number")
if (sparsity_factor <= 0) or (sparsity_factor >= 1):
raise Exception("Sparsity factor is within the range (0, 1)")
if gen_exp <= 0:
raise Exception("gen_exp is a positive number")
if (init_keep_rate <= 0) or (init_keep_rate > 1):
raise Exception("init_keep_rate is within the range (0, 1)")
rel, sex, _ = simulate_tree(
sample_size, sparsity_factor, gen_exp, init_keep_rate
)
write_fam(os.path.join(output_folder, "rel.fam"), rel, sex, None)
# if no subset of interest has been specified, keep all indices
if interest_in_relevant is None:
interest_in_relevant = np.ones((rel.shape[0])).astype(np.bool)
if ibd_path:
ibd = load_sparse_csr(os.path.join(output_folder, "IBD.npz"))
elif ibd:
if rel is None:
raise Exception("No relationship matrix given")
ibd, L, D = simple_numerator(rel)
# keep the original L and D because they are useless otherwise
save_sparse_csr(os.path.join(output_folder, "IBD.npz"), ibd)
save_sparse_csr(os.path.join(output_folder, "L.npz"), L)
save_sparse_csr(os.path.join(output_folder, "D.npz"), D)
else:
ibd = None
if epis_path:
epis = load_sparse_csr(os.path.join(output_folder, "Epistasis.npz"))
elif epis:
if ibd is None:
raise Exception("Pairwise-epistasis requires an ibd matrix")
epis = pairwise_epistasis(ibd)
save_sparse_csr(os.path.join(output_folder, "Epistasis.npz"), epis)
else:
epis = None
if dom_path:
dom = load_sparse_csr(os.path.join(output_folder, "Dominance.npz"))
elif dom:
if ibd is None or rel is None:
raise Exception(
"Dominance requires both an ibd matrix and a relationship matrix"
)
dom = dominance(rel, ibd)
save_sparse_csr(os.path.join(output_folder, "Dominance.npz"), dom)
else:
dom = None
covariance_matrices = []
for mat in [ibd, epis, dom]:
if mat is not None:
covariance_matrices.append(mat)
if cov is not None:
cov = np.hstack((cov, np.load(cov)))
else:
cov = sex[:, np.newaxis]
y = None
if gen_y:
sigs = np.random.rand(len(covariance_matrices) + 1)
sigs /= sigs.sum()
fe = np.random.rand(cov.shape[1] + intercept) / 100
print(
"Generating y with fixed effects: {} and sigmas : {}".format(
fe, sigs
)
)
y = simulate_phenotype(covariance_matrices, cov, sigs, fe, intercept)
np.save(os.path.join(output_folder, "y.npy"), y)
if y is not None:
y = np.load(y)
if he:
print(compute_HE(y, cov, covariance_matrices, intercept))
if lmm:
print(
LMM(
SparseCholesky(),
covariance_matrices,
cov,
y,
with_intercept=intercept,
reml=reml,
sim_num=sim_num,
)
)
help='Gen size = gen_exp X prev gen size')
parser.add_argument(
'--init_keep_rate',
dest='init_keep_rate',
type=float,
default=0.8,
help='1 - number of edges to remove before iteration begins')
parser.add_argument('--save_folder',
dest='save_folder',
type=str,
default='.',
help='which folder it should save the output to.')
args = parser.parse_args()
if args.sample_size <= 0:
raise Exception("Sample size should be a positive number")
if (args.sparsity_factor <= 0) or (args.sparsity_factor >= 1):
raise Exception("Sparsity factor is within the range (0, 1)")
if args.gen_exp <= 0:
raise Exception("gen_exp is a positive number")
if (args.init_keep_rate <= 0) or (args.init_keep_rate > 1):
raise Exception("init_keep_rate is within the range (0, 1)")
rel = load_sparse_csr('rel.npz')
print(count_IBD_nonzero(rel))
# rel, sex, gen_ind = simulate_tree(args.sample_size, args.sparsity_factor, args.gen_exp, args.init_keep_rate)
# save_sparse_csr(os.path.join(args.save_folder, 'rel.npz'), rel)
# np.save(os.path.join(args.save_folder, 'sex.npy'), sex)
# np.save(os.path.join(args.save_folder, 'gen_ind.npy'), gen_ind)
dest="init_keep_rate",
type=float,
default=0.8,
help="1 - number of edges to remove before iteration begins",
)
parser.add_argument(
"--save_folder",
dest="save_folder",
type=str,
default=".",
help="which folder it should save the output to.",
)
args = parser.parse_args()
if args.sample_size <= 0:
raise Exception("Sample size should be a positive number")
if (args.sparsity_factor <= 0) or (args.sparsity_factor >= 1):
raise Exception("Sparsity factor is within the range (0, 1)")
if args.gen_exp <= 0:
raise Exception("gen_exp is a positive number")
if (args.init_keep_rate <= 0) or (args.init_keep_rate > 1):
raise Exception("init_keep_rate is within the range (0, 1)")
rel = load_sparse_csr("rel.npz")
print(count_IBD_nonzero(rel))
# rel, sex, gen_ind = simulate_tree(args.sample_size, args.sparsity_factor, args.gen_exp, args.init_keep_rate)
# save_sparse_csr(os.path.join(args.save_folder, 'rel.npz'), rel)
# np.save(os.path.join(args.save_folder, 'sex.npy'), sex)
# np.save(os.path.join(args.save_folder, 'gen_ind.npy'), gen_ind)