def __init__(
self,
table: biom.table.Table,
metadata: pd.DataFrame,
model_path: str,
num_iter: int = 500,
num_warmup: int = None,
chains: int = 4,
seed: float = 42,
parallelize_across: str = "chains"
):
self.table = table
self.num_iter = num_iter
if num_warmup is None:
self.num_warmup = num_iter
else:
self.num_warmup = num_warmup
self.chains = chains
self.seed = seed
self.feature_names = table.ids(axis="observation")
self.sample_names = table.ids(axis="sample")
self.model_path = model_path
self.sm = None
self.fit = None
self.parallelize_across = parallelize_across
self.dat = {
"y": table.matrix_data.todense().T.astype(int),
"D": table.shape[0], # number of features
"N": table.shape[1], # number of samples
}
self.specifications = dict()
def __init__(
self,
table: biom.table.Table,
model_path: str,
num_iter: int = 500,
num_warmup: int = None,
chains: int = 4,
seed: float = 42,
):
self.num_iter = num_iter
if num_warmup is None:
self.num_warmup = num_iter
else:
self.num_warmup = num_warmup
self.chains = chains
self.seed = seed
self.sample_names = table.ids(axis="sample")
self.model_path = model_path
self.sm = None
self.fit = None
self.dat = {
"D": table.shape[0], # number of features
"N": table.shape[1], # number of samples
}
self.specified = False
def __init__(self,
table: biom.table.Table,
formula: str,
group_var: str,
metadata: pd.DataFrame,
num_iter: int = 500,
num_warmup: int = None,
chains: int = 4,
seed: float = 42,
beta_prior: float = 5.0,
cauchy_scale: float = 5.0,
group_var_prior: float = 1.0):
filepath = DEFAULT_MODEL_DICT["negative_binomial"]["lme"]
super().__init__(table=table,
formula=formula,
metadata=metadata,
model_path=filepath,
num_iter=num_iter,
num_warmup=num_warmup,
chains=chains,
seed=seed,
parallelize_across="chains")
# Encode group IDs starting at 1 because Stan 1-indexes arrays
group_var_series = metadata[group_var].loc[self.sample_names]
samp_subj_map = group_var_series.astype("category").cat.codes + 1
# Encoding as categories uses alphabetic sorting
self.groups = np.sort(group_var_series.unique())
param_dict = {
"depth": np.log(table.sum(axis="sample")), # sampling depths
"B_p": beta_prior,
"phi_s": cauchy_scale,
"S": len(group_var_series.unique()),
"subj_ids": samp_subj_map.values,
"u_p": group_var_prior
}
self.add_parameters(param_dict)
self.specify_model(params=["beta", "phi", "subj_int"],
dims={
"beta": ["covariate", "feature"],
"phi": ["feature"],
"subj_int": ["group", "feature"],
"log_lhood": ["tbl_sample", "feature"],
"y_predict": ["tbl_sample", "feature"]
},
coords={
"covariate": self.colnames,
"feature": self.feature_names,
"tbl_sample": self.sample_names,
"group": self.groups
},
include_observed_data=True,
posterior_predictive="y_predict",
log_likelihood="log_lhood")
if self.parallelize_across == "chains":
self.specifications["alr_params"] = ["beta", "subj_int"]
def __init__(
self,
table: biom.table.Table,
formula: str,
metadata: pd.DataFrame,
num_iter: int = 500,
num_warmup: int = None,
chains: int = 4,
seed: float = 42,
beta_prior: float = 5.0,
cauchy_scale: float = 5.0,
parallelize_across: str = "chains",
):
filepath = DEFAULT_MODEL_DICT["negative_binomial"][parallelize_across]
super().__init__(table=table,
formula=formula,
metadata=metadata,
model_path=filepath,
num_iter=num_iter,
num_warmup=num_warmup,
chains=chains,
seed=seed,
parallelize_across=parallelize_across)
param_dict = {
"depth": np.log(table.sum(axis="sample")), # sampling depths
"B_p": beta_prior,
"phi_s": cauchy_scale
}
self.add_parameters(param_dict)
self.specify_model(params=["beta", "phi"],
dims={
"beta": ["covariate", "feature"],
"phi": ["feature"],
"log_lhood": ["tbl_sample", "feature"],
"y_predict": ["tbl_sample", "feature"]
},
coords={
"covariate": self.colnames,
"feature": self.feature_names,
"tbl_sample": self.sample_names
},
include_observed_data=True,
posterior_predictive="y_predict",
log_likelihood="log_lhood")
if self.parallelize_across == "chains":
self.specifications["alr_params"] = ["beta"]
def __init__(
self,
table: biom.table.Table,
formula: str,
metadata: pd.DataFrame,
num_iter: int = 500,
num_warmup: int = None,
chains: int = 4,
seed: float = 42,
beta_prior: float = 5.0,
):
filepath = DEFAULT_MODEL_DICT["multinomial"]
super().__init__(
table=table,
model_path=filepath,
num_iter=num_iter,
num_warmup=num_warmup,
chains=chains,
seed=seed,
)
self.create_regression(formula=formula, metadata=metadata)
param_dict = {
"B_p": beta_prior,
"depth": table.sum(axis="sample").astype(int)
}
self.add_parameters(param_dict)
self.specify_model(params=["beta"],
dims={
"beta": ["covariate", "feature_alr"],
"log_lhood": ["tbl_sample"],
"y_predict": ["tbl_sample", "feature"]
},
coords={
"covariate": self.colnames,
"feature": self.feature_names,
"feature_alr": self.feature_names[1:],
"tbl_sample": self.sample_names,
},
include_observed_data=True,
posterior_predictive="y_predict",
log_likelihood="log_lhood")