Analysis Patterns

Most AnnNet workflows fall into a small number of practical patterns.

Multilayer Analysis

Use this when layers are part of the scientific model, for example modalities, states, or timepoints.

• Define aspects and layer tuples.
• Add intra-layer, inter-layer, or coupling edges.
• Analyze through AnnNet's multilayer operators or through a backend accessor.

Related pages: Multilayer and multi-aspect graphs, Layers, and Graph.

Notebooks: Multilayer Systems Biology and HEK293 multilayer network.

Hyperedges and Stoichiometry

Use this when binary edges are too limiting and you need directed hyperedges, weighted memberships, or SBML-style reaction structure.

• Import or construct hyperedges.
• Preserve endpoint-specific coefficients when needed.
• Export through a format that can preserve the structure, or choose an explicit projection when moving to simpler graph tools.

Related pages: Incidence representation, Graph, and SBML.

Notebook: AnnNet Showcase.

Slices and Scenario Management

Use this when you need multiple contexts in one container, such as train/eval splits, perturbation scenarios, or condition-specific subgraphs.

• Define slices.
• Switch the active slice when constructing or inspecting context-specific state.
• Compare memberships, weights, or annotations across slices.

Related pages: Slices and views, Slices, and Views.

Interoperability-First Analysis

Use this when AnnNet is the main container and you want to move into other tools without losing track of AnnNet-specific structure.

• Use G.nx, G.ig, or G.gt for graph-owned backend algorithm dispatch.
• Use annnet.adapters.to_* and annnet.adapters.from_* for explicit in-memory conversion.
• Use the native .annnet format when AnnNet should remain the system of record.

Related pages: Interoperability, Backend accessors, and Storage and IO.