EpigeneticPacemaker
Parameters
iter_limit: int, default=100
The maximum number of iterations
error_tolerance: float, default=0.00001
Optimization tolerance, if the error between two consecutive iterations is below the tolerance threshold the model will stop fitting
Attributes
EPM: Dict[model_parameters]
- MC_error - initial model error after first model iteration
- MC_rates - initial site rates after first model iteration
- MC_intercepts - initial sites intercepts after first model iteration
- MC_rss - initial residual sum of squares
- EPM_error - observed fit model error
- EPM_rates - fit model rates
- EPM_intercepts - fit model intercepts
- EPM_iter - number of iteration used to fit model
- EPM_rss - fit model residual of sum squares
- chi2 - log(MC_rss / EPM_rss) * number_sites * number_samples
- pval - chi2 p
Methods
fit(meth_array: numpy.array, states: numpy.array)
Fit an EPM model with an m x n methylation array and n methylation states, where m is the number of methylation sites and n is the number of samples. The meth_array and states should be passed an numpy arrays or an assertion error will be thrown.
predict(meth_array: numpy.array) -> predicted epigenetic states: numpy.array
Predict n epigenetic states given an m x n methylation array
score(meth_array: numpy.array, states: numpy.array) -> Pearson R: Tuple[significance, R value]
Predict n epigenetic states given an m x n methylation array and compare against known values. Note, predicted values may not be linearly dependent on input values.
EpigeneticPacemakerCV
Parameters
iter_limit: int, default=100
The maximum number of iterations
error_tolerance: float, default=0.00001
Optimization tolerance, if the error between two consecutive iterations is below the tolerance threshold the model will stop fitting
verbose: bool, default=False
Verbose model fitting
cv_folds: int, default=3
Number of cross validation folds
randomize_order: bool, default=True
Randomize the test folds, otherwise step through the CV folds in order
Attributes
EPM: Dict[model_parameters]
- EPM_rates - CV model rates
- EPM_intercepts - CV model intercepts
models: Dict[cv models]
Cross validation model attributes for each fold
predicted_states: numpy.ndarray Predicted epigenetic state for test samples, ie. samples left out of model fitting for a particular fold
Methods
fit(meth_array: numpy.array, states: numpy.ndarray)
Fit an EPM model with an (m \times n) methylation array and n methylation states, where m is the number of methylation sites and n is the number of samples. The meth_array and states should be passed an numpy arrays or an assertion error will be thrown.
predict(meth_array: numpy.array) -> predicted epigentic states: numpy.ndarray
Predict n epigenetic states given an m x n methylation array
score(meth_array: numpy.ndarray, states: numpy.ndarray) -> Pearson R: Tuple[significance, R value]
Predict n epigenetic states given an m x n methylation array and compare against known values. Note, predicted values may not be linearly dependent on input values.