NATURA-EVIDENCE

A computational pipeline for natural product drug discovery evidence generation. Searches external scientific databases, predicts drug-target interactions, performs molecular docking, ADMET analysis, and generates comprehensive evidence dossiers.

Quick Start

bash run.sh

This starts all services:

• Frontend: http://localhost:5173
• API: http://localhost:8000
• API Docs: http://localhost:8000/docs

Architecture

User creates Product (name, category)
    |
    v
Pipeline (Celery DAG, 16 steps)
    |
    +-- Compound Discovery (PubChem, ChEBI, NPASS, FooDB, Dr Duke, KNApSAcK, Wikidata, Literature)
    +-- Target Prediction (Ensemble ML + SwissTargetPrediction)
    +-- Protein Retrieval (UniProt, PDB, AlphaFold)
    +-- Disease Association (OpenTargets, CTD, DisGeNET)
    +-- Pathway Analysis (KEGG, Reactome, Gene Ontology)
    +-- Network Pharmacology (STRING PPI, NetworkX)
    +-- Molecular Docking (AutoDock Vina / RDKit fallback)
    +-- ADMET Prediction (Dual predictor + SwissADME)
    +-- Literature Mining (PubMed, EuropePMC, Semantic Scholar)
    +-- Clinical Trials (ClinicalTrials.gov)
    +-- AI Evidence Extraction (LLM-based)
    +-- Evidence Scoring & Hallucination Protection
    +-- Dossier Assembly (PDF, DOCX, HTML)
    |
    v
16-tab Evidence Dossier Viewer

Tech Stack

Layer	Technology
Backend	Python 3.12, FastAPI, SQLAlchemy 2.0 (async), Celery, Redis
Database	PostgreSQL 15 (or SQLite fallback)
Frontend	React 19, TypeScript, Vite, TailwindCSS, Recharts, Cytoscape.js
ML	RDKit, Chemprop, DeepChem, PyTorch
Docking	AutoDock Vina, RDKit conformer energy

Project Structure

NATURA-EVIDENCE/
├── run.sh                          # Single-command launcher
├── docker-compose.yml              # PostgreSQL + Redis
├── natura-evidence-api/            # Python backend
│   ├── app/
│   │   ├── main.py                 # FastAPI entry point
│   │   ├── config.py               # Settings (env vars)
│   │   ├── database.py             # SQLAlchemy async engine
│   │   ├── api/                    # 28 API routers
│   │   ├── core/                   # 61 core modules (ML, docking, etc.)
│   │   ├── models/                 # 14 database models
│   │   ├── services/               # 25 external API clients
│   │   └── tasks/                  # Celery pipeline tasks
│   └── tests/                      # 53 test files
|── natura-evidence-ui/             # React frontend
    └── src/
        ├── pages/                  # 8 page components
        ├── components/dossier/     # 16 tab components
        ├── lib/api.ts              # API client
        └── types/                  # TypeScript interfaces

Pipeline Steps

#	Step	API Sources	Output
1	Natural Product Resolution	Wikidata	Taxonomy, synonyms
2	Compound Discovery	PubChem, ChEBI, NPASS, FooDB, Dr Duke, KNApSAcK, Wikidata, Literature	71 compounds
3	Compound Enrichment	RDKit	Molecular descriptors, drug-likeness
4	Target Prediction	Ensemble ML + SwissTargetPrediction	400+ protein targets
5	Protein Retrieval	UniProt, PDB, AlphaFold	3D structures, sequences
6	Disease Association	OpenTargets, CTD	Gene-disease associations
7	Pathway Discovery	KEGG, Reactome, GO	Enriched pathways
8	Network Pharmacology	STRING, NetworkX	Compound-Target-Disease graph
9	Molecular Docking	AutoDock Vina / RDKit	Binding energies, poses
10	ADMET Prediction	Dual predictor + SwissADME	28 ADMET endpoints
11	Literature Mining	PubMed, EuropePMC	Scientific publications
12	Clinical Trials	ClinicalTrials.gov	Trial data
13	AI Evidence Extraction	LLM-based	Structured claims
14	Evidence Scoring	Quality assessment	Evidence tiers
15	Hallucination Protection	Cross-verification	Verified claims
16	Dossier Assembly	All sources	PDF, DOCX, HTML

API Endpoints

Endpoint	Purpose
POST /api/pipeline/{id}/start	Start pipeline
GET /api/pipeline/{id}/status	Pipeline status + SSE stream
GET /api/products/{id}/compounds	Compound list
GET /api/products/{id}/targets	Target predictions
GET /api/products/{id}/diseases	Disease associations
GET /api/products/{id}/pathways	Pathway analysis
GET /api/products/{id}/network	Network graph
GET /api/products/{id}/docking	Docking results
GET /api/products/{id}/admet	ADMET profiles
GET /api/products/{id}/evidence	Literature evidence
GET /api/products/{id}/clinical-trials	Clinical trials
GET /api/products/{id}/proteins	Protein structures

Configuration

Environment variables (in .env):

# Database
DATABASE_URL=sqlite+aiosqlite:///./natura_evidence.db

# Redis
REDIS_URL=redis://localhost:6380/0

# Celery
CELERY_BROKER_URL=redis://localhost:6380/0

# API Keys (optional — enhances data quality)
DISGENET_API_KEY=
NCBI_API_KEY=
PUBCHEM_API_KEY=

Running

# Start everything
bash run.sh

# Or start individually
cd natura-evidence-api
source .venv/bin/activate
uvicorn app.main:app --host 0.0.0.0 --port 8000

celery -A app.tasks.celery_app worker -Q default,compound,target,protein,disease,pathway,network,docking,admet,literature,clinical,ai,dossier --pool=solo --concurrency=1

cd ../natura-evidence-ui
npm run dev

Docker (Infrastructure Only)

docker-compose up -d
# Starts PostgreSQL (port 5433) and Redis (port 6380)

Testing

cd natura-evidence-api
source .venv/bin/activate
pytest tests/ -v

License

Private — for research use only.