pygot.tools.analysis.ProbabilityModel

class pygot.tools.analysis.ProbabilityModel(device=None)

Bases: object

Probability model for pseudotime estimation

Example

---

#Assume the velocity are already fitted in pca space
embedding_key = 'X_pca'
velocity_key = 'velocity_pca'

# Fit the probability model
pm = pygot.tl.analysis.ProbabilityModel()
history = pm.fit(adata,  embedding_key=embedding_key, velocity_key=velocity_key)

# Estimated pseudotime of cells
adata.obs['pseudotime'] = pm.estimate_pseudotime(adata) # pseudotime
adata.obs['var'] = pm.estimate_variance(adata) # variance of time

__init__(device=None)

Init model

Arguments:

device: device

torch device

Methods

__init__([device])	Init model
estimate_pseudotime(adata[, mode])	estimate the pseudotime
estimate_variance(adata)	estimate the variance of pseudotime
fit(adata, embedding_key, velocity_key[, ...])	fit model
to(device)

estimate_pseudotime(adata, mode='mean')

estimate the pseudotime

Arguments:

adata: AnnData

Annotated data matrix.

returns:

:math:`t^*|x` – pseudotime of cells

rtype:

ndarray

estimate_variance(adata)

estimate the variance of pseudotime

Arguments:

adata: AnnData

Annotated data matrix.

returns:

var – variance of cell time

rtype:

ndarray

fit(adata, embedding_key, velocity_key, n_neighbors=30, n_iters=500, mini_batch=True, batch_size=512)

fit model

Arguments:

adata: AnnData

Annotated data matrix.

embedding_key: str (default: None)

Name of latent space, in adata.obsm.

velocity: str (default: None)

Name of latent velocity, in adata.obsm. Use to do variantional inference of conditonal time distribution if it offers.

time_key: str (default: None)

Name of time label, in adata.obs. Use as addition information for conditonal time distribution fitting if it offers.

n_neighbors: int (default: 30)

Number of neighbors of cell

n_iters: float (default: 500)

Number of training iterations

mini_batch: bool (default: True)

Use mini-batch training or not

batch_size: int (default: 512)

Number of batch size

to(device)