API

This section provides documentation for the fava.py module.

# List the members in the order you want them to appear :members:

VAE cook pairs_after_cutoff

class fava.VAE(*args, **kwargs)

Bases: Model

Variational Autoencoder model class.

Parameters:

  • opt (tf.keras.optimizers.Optimizer) – Optimizer for the model.

  • x_train (np.ndarray) – Training data.

  • x_test (np.ndarray) – Test data.

  • batch_size (int) – Batch size for training.

  • original_dim (int) – Dimension of the input data.

  • hidden_layer (int) – Number of units in the hidden layer.

  • latent_dim (int) – Dimension of the latent space.

  • epochs (int) – Number of training epochs.

fava.cook(data, log2_normalization=True, hidden_layer=None, latent_dim=None, epochs=50, batch_size=32, interaction_count=100000, correlation_type='pearson', CC_cutoff=None)

Preprocess data, train a Variational Autoencoder (VAE), and create filtered protein pairs.

Parameters:

  • data (np.ndarray or anndata._core.anndata.AnnData) – Input data or AnnData object.

  • log2_normalization (bool, optional) – Whether to apply log2 normalization, by default True.

  • hidden_layer (int, optional) – Number of units in the hidden layer, by default None.

  • latent_dim (int, optional) – Dimension of the latent space, by default None.

  • epochs (int, optional) – Number of training epochs, by default 50.

  • batch_size (int, optional) – Batch size for training, by default 32.

  • interaction_count (int, optional) – Maximum number of interactions to include, by default 100000.

  • correlation_type (str, optional) – Type of correlation to use (Pearson or Spearman), by default Pearson.

  • CC_cutoff (float, optional) – Correlation Coefficient cutoff, by default None.

Returns:

final_pairs – Filtered protein pairs based on correlation and cutoffs.

Return type:

pd.DataFrame