regvelo.tools.TFscreening

regvelo.tools.TFscreening(adata, prior_graph, lam=1, lam2=0, soft_constraint=True, TF_list=None, cluster_label=None, terminal_states=None, terminal_states_manual=None, KO_list=None, n_states=8, cutoff=0.001, max_nruns=5, method='likelihood', dir=None)[source]

Perform repeated in silico TF regulon knock-out screening.

Parameters:

  • adata (AnnData) – Annotated data matrix.

  • prior_graph (Tensor) – A prior graph for RegVelo inference.

  • lam (int) – Regularization parameter for controling the strengths of adding prior knowledge.

  • lam2 (int) – Regularization parameter for controling the strengths of L1 regularization to the Jacobian matrix.

  • soft_constraint (bool) – Apply soft constraint mode RegVelo.

  • TF_list (Union[str, list[str], dict[str, list[str]], Series, None]) – The TF list used for RegVelo inference.

  • cluster_label (Optional[str]) – Key in adata.obs to associate names and colors with terminal_states.

  • terminal_states (Union[str, list[str], dict[str, list[str]], Series, None]) – subset of macrostates.

  • KO_list (Union[str, list[str], dict[str, list[str]], Series, None]) – List of TF names or combinations (e.g., [“geneA”, “geneB_geneC”]).

  • n_states (int | list[int]) – Number of macrostates to compute.

  • cutoff (float) – Threshold to zero out TF-target links during knock-out.

  • max_nruns (float) – Maximum number of runs, soft constrainted RegVelo model need to have repeat runs to get stable perturbation results. Set to 1 if soft_constraint=False.

  • method (str) – Method to quantify perturbation effects

  • dir (Optional[str]) – Directory to store intermediate model files and results.

  • terminal_states_manual (dict | None)

Return type:

tuple[DataFrame, DataFrame]

Returns:

:

  • DataFrame containing the coefficients of the perturbation effects.

  • DataFrame containing the p-values of the perturbation effects.