pygot.tools.analysis.GRN

class pygot.tools.analysis.GRN

Bases: object

Gene regulatory network infered by velocity linear regression

Example:

grn = GRN()
grn_adata = grn.fit(adata, species='human')
print(grn_adata.ranked_edges.head()) #print the top regulatory relationship

__init__()

Methods

__init__()
fit(adata[, TF_constrain, TF_names, ...])	fit the gene regulatory network

fit(adata, TF_constrain=True, TF_names=None, species='human', non_negative=True, layer_key=None, n_epoch=10000, lr=0.01, l1_penalty=0.005, init_beta=1.0, min_beta=1.0, init_jacobian=None, early_stopping=True, batch_size=2048, val_split=0.2, device=None, lineage_key=None)

fit the gene regulatory network

Arguments:

adata: AnnData

Annotated data matrix, gene velocity should be stored in adata.layers[‘velocity’]

TF_constrain: bool (default: True)

Only fit the transcriptional factor(TF)

TF_names: list (default: None)

Names of TF, if None, use default TF names

species: ‘human’ or ‘mm’ (default: ‘human’)

Default TF names of species

non_negative: bool (default: True)

ONLY fit positive regulatory relationship, which may avoid overfit

layer_key: str (default: None)

Data use as x, if None, use adata.X else should stored in adata.layers

n_epoch: int (default: 10000)

Number of training epochs

lr: float (default: 0.01)

Learning rate

l1_penalty: float (default: 0.005)

l1 weight, control sparsity of grn

init_beta: float or GeneDegradation (default: 1.0)

Initial gene degrade rate

min_beta: float (default: 1.0)

Lower bound of degrade rate

init_jacobian: np.ndarray (default: None)

Initial grn

early_stopping: bool (default: True)

Early stopping training

batch_size: int (default: 2048)

Batch size of mini-batch training

val_split: float (default: 0.2)

Validation dataset portion

device: torch.device (default: None)

torch device

lineage_key: discard

This parameter is discarded

returns:

grn_data – gene regulatory network

rtype:

GRNData