Deterministic, Explainable Materials Discovery Knowledge Graph for Scientific Exploration
MaterialGraph is an open-source platform for deterministic, explainable materials discovery and scientific decision support. It combines graph-based knowledge representation, explainable scoring, graph analytics, and research-oriented exploration to help researchers investigate scientifically plausible material alternatives.
Unlike autonomous AI systems, MaterialGraph does not replace scientific judgment. It computes, ranks, explains, and contextualizes research opportunities while keeping researchers in control of scientific decisions.
Modern materials research requires balancing chemistry, stability, criticality, supply risk, and scientific plausibility.
MaterialGraph helps researchers:
- Discover scientifically related materials
- Explore explainable substitution pathways
- Analyze graph relationships and communities
- Evaluate research objectives
- Understand risks, trade-offs, and assumptions
- Make informed scientific decisions
Additional project documentation is available in the docs/ directory.
| Document | Description |
|---|---|
| Getting Started | Local development setup and project bootstrapping |
| System Architecture | Current architecture and intelligence layer design |
| Scientific Principles | Scientific principles and design rationale |
| Research Architecture | Research-focused architecture and design decisions |
| Roadmap | Future development plans and feature roadmap |
| Known Issues | Current limitations and tracked issues |
| Deployment Guide | Production deployment using AWS EC2, Neon PostgreSQL, systemd, and Nginx |
- Deterministic reasoning
- Explainable intelligence
- Graph-driven scientific exploration
- Researcher-in-the-loop decision support
- Rank, explain, warn, and score
- No LLM reasoning in scientific computation
- Material Graph Foundation
- Material Neighborhood Intelligence
- Material Family Intelligence
- Similarity Engine
- Recommendation Engine
- Criticality Analysis
- Scenario Policy Engine
- Discovery Candidate Engine
- Explainable Discovery Scoring
- Discovery Warnings
- Substitution Path Engine
- Multi-Hop Discovery Chains
- Discovery Path Ranking
- Research Objective Exploration
- Graph Builder
- Graph Traversal
- BFS / DFS / Dijkstra / K-shortest Paths
- Community Detection
- Community Intelligence
- Ranked Subgraph Exploration
- Graph Analytics
- Material Quality
- Node & Edge Intelligence
Materials Project
│
▼
Material Graph Foundation
│
▼
Foundation Intelligence
│
▼
Discovery Intelligence
│
▼
Knowledge Graph Intelligence
│
▼
Research Intelligence
│
▼
Scientific Knowledge Layer (Future)
- Python
- FastAPI
- SQLAlchemy
- PostgreSQL
- Alembic
- NetworkX
- Pydantic v2
- AWS EC2
- Nginx
- systemd
- Docker
- pytest
git clone https://github.com/<username>/materialgraph.git
cd materialgraph
python -m venv .venv
pip install -r requirements.txt
alembic upgrade head
python scripts/import_materials_project.py
uvicorn app.main:app --reloadSee the docs/ directory for:
- System Architecture
- Scientific Principles
- Getting Started
- Deployment Guide
- Technical Notes
- Roadmap
- Multi-element constraints
- Application-aware exploration
- USGS criticality enrichment
- Geopolitical, toxicity, and recyclability policies
Completed:
- Community Detection
- Community Intelligence
- Ranked Subgraph Exploration
- Research Objective Exploration
Current Focus:
- Scientific Pathway Analysis
- Research Opportunity Analysis
Future:
- Research Gap Analysis
- Hypothesis Exploration
- Multi-objective Optimization
- PostgreSQL graph jobs
- Go GraphCompute Worker
- Background analytics
- Rust graph engine
- Large-scale traversal
- High-performance scientific path search
MaterialGraph assists scientific exploration. It does not:
- Replace DFT calculations
- Guarantee synthesis feasibility
- Replace laboratory validation
- Replace scientific judgment
Researchers remain responsible for evaluating, selecting, and validating research opportunities.
MIT License