Initialization

CRISGI.init

Function

__init__(
    adata,
    bg_net=None,
    bg_net_score_cutoff=850,
    genes=None,
    n_hvg=5000,
    n_pcs=30,
    interactions=None,
    n_threads=5,
    interaction_methods=["prod"],
    organism="human",
    class_type="time",
    dataset="test",
    out_dir="./out",
)

Initializes a CRISGI object for single-cell gene interaction analysis. Sets up the AnnData object, prepares the background network, and configures preprocessing and output directories.

Parameters

Name	Type	Description
adata	AnnData	The annotated data matrix (cells x genes) to be analyzed.
bg_net	Optional	Precomputed background network (default: None).
bg_net_score_cutoff	int	Score cutoff for background network edges (default: 850).
genes	list or None	List of gene names to include (default: None, uses highly variable genes if available).
n_hvg	int or None	Number of highly variable genes to select (default: 5000).
n_pcs	int	Number of principal components for PCA (default: 30).
interactions	Optional	Predefined gene interactions (default: None).
n_threads	int	Number of threads to use for computation (default: 5).
interaction_methods	list	Methods for interaction inference (default: ["prod"]).
organism	str	Organism name (default: 'human').
class_type	str	Type of analysis class (default: 'time').
dataset	str	Dataset identifier (default: 'test').
out_dir	str	Output directory for results (default: './out').

Return type

None

Returns

Initializes the CRISGI object and prepares it for downstream analysis.

Attributes Set

• adata: Processed AnnData object.
• interaction_methods: List of interaction inference methods.
• organism: Organism name.
• n_threads: Number of threads for computation.
• dataset: Dataset identifier.
• class_type: Type of analysis class.
• out_dir: Output directory path.
• bg_net_score_cutoff: Score cutoff for background network.
• adata.varm['bg_net']: Background network matrix.

Example

import scanpy as sc
from crisgi import CRISGI

adata = sc.read_h5ad('example_data.h5ad')
crisgi = CRISGI(
    adata,
    interaction_methods=["prod"],
    n_hvg=3000,
    n_pcs=20,
    organism='human',
    out_dir='./crisgi_results'
)

---

CRISGITime.init

Function

__init__(
    self,
    adata,
    device="cpu",
    model_type="cnn",
    ae_path=None,
    mlp_path=None,
    model_path=None,
    **kwargs
)

Initializes a CRISGITime object for time-series or temporal single-cell gene interaction analysis. Inherits from CRISGI and adds model selection and device configuration.

Parameters

Name	Type	Description
adata	AnnData	The annotated data matrix (cells x genes) to be analyzed.
device	str	Device to use for computation ('cpu' or 'cuda', default: 'cpu').
model_type	str	Model type to use ('cnn', 'simple_cnn', 'logistic', default: 'cnn').
ae_path	str/None	Path to autoencoder model weights (optional).
mlp_path	str/None	Path to MLP model weights (optional).
model_path	str/None	Path to logistic model weights (optional).
**kwargs	dict	Additional keyword arguments passed to CRISGI.init.

Return type

None

Returns

Initializes the CRISGITime object and sets up the selected model for downstream analysis.

Attributes Set

• All attributes from CRISGI.
• device: Computation device.
• model_type: Selected model type.
• model: Instantiated model object.

Example

import scanpy as sc
from crisgi import CRISGITime

adata = sc.read_h5ad('example_data.h5ad')
crisgi_time = CRISGITime(
    adata,
    interaction_methods=["prod"],
    device='cuda',
    model_type='cnn',
    out_dir='./crisgi_time_results'
)