ANDREA

Automated Network Discovery, Reproducible Evaluation and Analysis

ANDREA is a catalog-driven platform for generating benchmark expression datasets, running gene regulatory network (GRN) inference tools, evaluating inferred networks against reference truth and comparing networks across tools, contexts and runs. It exposes the same workflows through a command-line interface and local browser GUIs, while keeping public artifacts as plain JSON, CSV, SQLite and ZIP files.

What ANDREA Provides

• A standardized catalog of simulator and inference-tool integrations, with compatibility rules, parameter schemas, Docker wrappers and provenance metadata.
• Four interoperable workflows: generate-data -> infer-network -> evaluate-inference -> compare-networks.
• Strict handoff bundles so outputs from one command can be consumed directly by the next one.
• Context-aware handling of global, group-level and column-level network outputs, including emulated and aggregated execution strategies where appropriate.
• Cost-aware planning and execution orchestration for Dockerized simulator and inference wrappers.
• Evaluation reports and interactive comparison views for distance maps, metric overlays and edge-level differences.
• Support for temporary external Docker inference tools that follow ANDREA's standard container contract.

Installation

Install the release package:

pip install ANDREA

For a development checkout:

git clone https://github.com/AdrianSeguraOrtiz/ANDREA.git
cd ANDREA
python -m pip install -e ".[dev]"

Docker is required for workflows that execute simulator or inference wrappers. See docs/installation.md for the full setup notes.

Quick Start

Inspect the CLI:

andrea --help
andrea generate-data --help
andrea infer-network --help
andrea evaluate-inference --help
andrea compare-networks --help

Open the local GUIs:

andrea gui generate-data
andrea gui infer-network
andrea gui evaluate-inference
andrea gui compare-networks

The GUIs are local browser applications. ZIP bundles are mainly intended for GUI handoff; CLI and Python users can usually pass report JSON paths directly.

Workflows

See docs/workflows.md for the workflow contracts and docs/gui.md, docs/cli.md and docs/core.md for GUI, command-line and Python usage.

Documentation

• Installation
• Workflow contracts
• GUI guide
• CLI guide
• Core Python guide
• Catalogs and coverage
• External Docker tools
• Full documentation index