VegaSCVI

class vega.VegaSCVI(adata, gmt_paths=None, add_nodes=1, min_genes=0, max_genes=5000, positive_decoder=True, n_hidden=600, n_layers=2, gene_likelihood='zinb', dropout_rate=0.1, z_dropout=0, use_cuda=True, **model_kwargs)[source]

VEGA: VAE Enhanced by Gene Annotations [Seninge2021].

Parameters

  • adata – AnnData object that has been registered via setup_anndata().

  • gmt_paths – A single or list of paths to .GMT files with gene annotations for GMVs initialization.

  • add_nodes – Number of additional fully-connected decoder nodes (unannotated GMVs).

  • min_genes – Minimum gene size for GMVs.

  • max_genes – Maximum gene size for GMVs.

  • positive_decoder – Whether to constrain decoder to positive weights.

  • n_hidden – Number of nodes per hidden layer.

  • n_layers – Number of hidden layers used for encoder NN.

  • gene_likelihood – Likelihood function for the generative model.

  • dropout_rate – Dropout rate for neural networks.

  • use_cuda – Using GPU with CUDA