pygot.preprocessing.GS_VAE

class pygot.preprocessing.GS_VAE(device=None)

Bases: object

Dimension reduction model

Non-linear dimension reduction model (based on VAE)

This model contains normal VAE model and Gene Set augmented VAE model (GS-VAE)

Example:

To use VAE:

vae = pygot.pp.gs_vae()
vae.register_model(adata, latent_dim=10)
adata.obsm['X_latent'] = vae.fit_transform(adata)

To use GS-VAE:

gs_vae = pygot.pp.gs_vae()
auc_space = gs_vae.precompute_gs(adata, gene_set_book_paths, mouse=False) # compute gene set score by AUCell
adata = gs_vae.process_gs(adata, auc_space) # process gene set score into adata
gs_vae.register_model(adata, latent_dim=10)
adata.obsm['X_latent'] = gs_vae.fit_transform(adata)

denoised_gs = gs_vae.pred_gs_using_z(adata.obsm['X_latent']) # predict denoised gene set score

Warning

The GS-VAE is not suggested to use because it still under developing.

__init__(device=None)

Methods

__init__([device])
fit(adata[, lr, batch_size, n_epoch])
fit_transform(adata[, lr, batch_size, n_epoch])
inverse_transform(x_latent)
precompute_gs(adata, gene_set_book_paths[, ...])
pred_gs_using_x(x)
pred_gs_using_z(x_latent)
process_gs(adata, auc_space[, c_filter, ...])
register_model(adata, latent_dim[, lam, ...])
transform(adata)

fit(adata, lr=0.0001, batch_size=128, n_epoch=100)

fit_transform(adata, lr=0.0001, batch_size=128, n_epoch=100)

inverse_transform(x_latent)

precompute_gs(adata, gene_set_book_paths, mouse=False)

pred_gs_using_x(x)

pred_gs_using_z(x_latent)

process_gs(adata, auc_space, c_filter=False, c_threshold=0.7)

register_model(adata, latent_dim, lam=1, hinge_value=0.01, beta=1e-06)

transform(adata)