Genome-Scale Mechanistic Networks for Biological Pathway Analysis

Napistu is an open-source project for creating and mining genome-scale networks of cellular physiology. Build consensus pathway models from diverse data sources, analyze biological networks, and train graph neural networks to predict regulatory interactions.

• Unified Representation - Encode diverse data sources using the sbml_dfs structure to faithfully represent molecular biology and biochemistry
• Consensus Models - Aggregate complementary sources into comprehensive networks where high-quality but incomplete interactions are supported by more comprehensive yet speculative data sources
• Network Analysis - Translate pathway models into genome-scale graphical networks for propagation, search, and neighborhood discovery
• Machine Learning - Train graph neural networks on biological networks to predict regulatory interactions and pathway relationships
• Easy Access - Pre-processed pathway data available on Google Cloud Storage and HuggingFace; MCP server for AI agent integration

Core Python library for pathway representations and network-based searches.

pip install napistu

PyTorch Geometric framework for training GNNs on biological pathways.

pip install 'napistu-torch[pyg,lightning]'

R library for network visualization and utilities.

remotes::install_github("napistu/napistu-r")

• Website - Project landing page
• Tutorials - Examples and documentation
• Wiki - Core algorithms and data structures
• Blog - Deep dives and tutorials
• HuggingFace - Pre-trained models and datasets

• Napistu meets PyTorch Geometric - Predicting Regulatory Interactions with GNNs (Nov 2025)
• Napistu's Octopus: An 8-source human consensus pathway model (Oct 2025)
• Building AI-Friendly Scientific Software: A Model Context Protocol Journey (Sep 2025)

• Issues - GitHub Issues
• Chat - Visit napistu.com for interactive AI assistan