Network inference (corneto.methods)

Methods (corneto.methods)

This module provides the implementations of the various methods used in CORNETO. It is organized into several functional areas.

Signaling

Network methods to infer signaling networks from omics data.

CarnivalILP.build and CarnivalFlow.build accept explicit perturbation and transcription-factor mappings for one condition. Their build_many methods accept named conditions. milp_carnival remains as a compatibility formulation.

CellNOptDAG.build accepts binary input, measurement, and optional inhibitor mappings. build_many infers one shared connected reaction model while evaluating its Boolean state independently in every named condition.

LinearDAGDiscovery and signaling.CellNOptDAG are reusable mechanistic models. Both provide a thin fit(...) convenience wrapper that returns the same method instance, while retaining the composable build(...)/build_many(...) -> ProblemDef -> solve(...) lifecycle. After a usable solution has been obtained, predict(...) applies the fixed model to new externally specified conditions and evaluate(...) compares those predictions with supplied measurements. LinearDAGDiscovery also exposes residuals(...) for local structural-equation diagnostics: its forward loss and local loss have different semantics. CellNOptDAG evaluates its fixed selected reactions by deterministic Boolean propagation. Most other Method subclasses remain explanatory optimization methods; the base class does not expose an sklearn-style fit/predict API.

CarnivalFlow([lambda_reg, ...])	Flow-base, multi-sample CARNIVAL method for intracellular signaling.
CarnivalILP([beta_weight, max_dist, ...])	Multi-condition implementation of the CARNIVAL ILP formulation.
signaling.CellNOptDAG([lambda_reg, ...])	Infer a shared acyclic Boolean model from multiple conditions.
BidirectionalPHONEMeS([default_edge_cost, ...])	Infer one globally optimized network around regulated kinase anchors.
PHONEMeS([default_edge_cost, max_flow, ...])	Infer acyclic signaling networks from phosphoproteomic scores.
compute_phonemes_scores(pvalues, *[, ...])	Compute PHONEMeS node scores from p-values and fold changes.
milp_carnival(G, perturbations, measurements)	Build the supported single-condition CARNIVAL ILP formulation.

CellNOpt visualization

CellNOpt plotting utilities use the standard CORNETO graph renderers for network views and return Matplotlib figure/axes objects for data-fit views.

signaling.plot_cellnopt_model(method[, ...])	Plot a solved CellNOpt model using CORNETO's graph renderers.
signaling.plot_cellnopt_fit(method[, ...])	Compare CellNOpt measurements and predictions across conditions.

Causal discovery

LinearDAGDiscovery infers a sparse, acyclic linear structural model from continuous observational and perfect-intervention measurements. Candidate interactions come from a directed prior-knowledge network, and commodity flows ensure that selected edges are supported by intervention-to-response paths.

LinearDAGDiscovery([lambda_edges, ...])

Discover a sparse linear DAG inside a directed prior network.

CellNOpt and AnnNet

These helpers keep the signed network, perturbation conditions, and fitted CellNOptDAG results in one AnnNet object.

signaling.add_cellnopt_conditions(graph, *, ...)	Add CellNOpt perturbations and measurements as AnnNet condition layers.
signaling.build_cellnopt_from_annnet(method, ...)	Build a CellNOptDAG problem from a signed network and data in AnnNet.
signaling.add_cellnopt_results(graph, ...[, ...])	Add a solved CellNOptDAG model and its condition results to AnnNet.

Metabolism

Network methods for flux balance analysis in metabolic networks. Use build with explicit objectives, bounds, or expression values; build_from_data provides the advanced generic-data interface.

MultiSampleFBA([lambda_reg, beta_reg, ...])	Flux Balance Analysis (FBA) method for multiple samples.
MultiSampleIMAT([lambda_reg, beta_reg, eps, ...])	Integrative Metabolic Analysis Tool (iMAT) implementation for multiple samples.

Graph optimization

Methods for extracting optimal subnetworks.

PrizeCollectingSteinerTree([...])	Prize-Collecting Steiner Tree optimization method built on top of FlowMethod.
SteinerTreeFlow([max_flow, ...])	Basic Steiner Tree optimization method as a flow-based problem.
create_multisample_shortest_path(G, ...[, ...])	Build one shortest-path flow per source-target condition.
shortest_path(G, s, t[, edge_weights, ...])	Build a shortest-path optimization problem between two vertices.
solve_shortest_path(G, s, t[, edge_weights, ...])	Solve a shortest-path problem and return its selected edge indices.

Solution sampling

Utilities used by the indexed alternative-solution tutorials.

sampler.sample_alternative_solutions(...[, ...])

Generate alternative feasible solutions by perturbing one decision variable.