SciML Hub

SciML Hub is a unified platform for scientific machine learning, providing state-of-the-art Nexa models for predictions in biology (protein structure), astrophysics (stellar properties), and materials science (material properties). Access these models via a simple REST API, with results returned in JSON format including predictions and confidence scores (0-100%).

Quick Start

Prerequisites

• Python 3.9+
• An API key (sign up at scimlhub.com to get yours)
• requests library (pip install requests)

Example: Protein Structure Prediction

Predict the secondary structure of a protein sequence:

import requests

response = requests.post(
    "https://api.scimlhub.com/v1/bio/predict",
    headers={"X-API-Key": "your_api_key"},
    json={"sequence": "MAKQVKL"}
)

result = response.json()
print(result)
# Output: {"prediction": "H", "confidence": 80.56}

Example: Stellar Property Prediction

Estimate a star's mass:

response = requests.post(
    "https://api.scimlhub.com/v1/astro/predict",
    headers={"X-API-Key": "your_api_key"},
    json={
        "temp": 5778,  # Kelvin
        "luminosity": 1.0,  # Solar luminosity
        "metallicity": 0.0  # [Fe/H]
    }
)

result = response.json()
print(f"Stellar Mass: {result['prediction']} Solar masses")
print(f"Confidence: {result['confidence']}%")
# Output: {"prediction": 1.0, "confidence": 97.49}

Example: Material Property Prediction

Predict a material's band gap:

response = requests.post(
    "https://api.scimlhub.com/v1/materials/predict",
    headers={"X-API-Key": "your_api_key"},
    json={"structure": "POSCAR data string"}
)

result = response.json()
print(f"Band Gap: {result['prediction']} eV")
print(f"Confidence: {result['confidence']}%")
# Output: {"prediction": 2.5, "confidence": 98.5}

Core Models

Biology: HelixSynth-Pro (Protein Structure Prediction)

• Model: Variational Autoencoder (VAE) with diffusion
• Purpose: Predicts protein secondary structures (H: Helix, E: Sheet, C: Coil)
• Accuracy: 70.82% overall (Q3 score)
• Latency: ~78ms
• Details: See helixsynth-pro.ipynb and Technical_Whitepaper/analysis_results_20250219_202508.txt

Astrophysics: Stellar Classification

• Model: Ensemble (Random Forest, CatBoost, Neural Network)
• Purpose: Predicts stellar properties (mass, class: QSO, GALAXY, STAR)
• Accuracy: 97.49% on validation set
• Latency: ~45ms
• Details: See Technical_Whitepaper/Astrophysics results.txt

Materials Science: Materials GNN

• Model: Graph Neural Network (GNN)
• Purpose: Predicts material properties (band gap, formation energy)
• Accuracy: 98.5% on crystal structures
• Latency: ~62ms
• Details: See Material Science/generated_structures.csv

API Usage

The API endpoints return predictions and confidence scores in JSON format: {"prediction": value, "confidence": percentage}.

Endpoints

1. /v1/bio/predict - Protein Structure Prediction

• Method: POST
• Input:

  {
    "sequence": "MAKQVKL"  // Amino acid sequence (up to 1000 residues)
  }

• Output:

  {
    "prediction": "H",  // H (Helix), E (Sheet), C (Coil)
    "confidence": 80.56  // Confidence in percentage (0-100)
  }

2. /v1/astro/predict - Stellar Property Prediction

• Method: POST
• Input:

  {
    "temp": 5778,        // Temperature in Kelvin
    "luminosity": 1.0,   // Luminosity in solar units
    "metallicity": 0.0   // Metallicity [Fe/H]
  }

• Output:

  {
    "prediction": 1.0,   // Mass in solar masses
    "confidence": 97.49  // Confidence in percentage (0-100)
  }

3. /v1/materials/predict - Material Property Prediction

• Method: POST
• Input:

  {
    "structure": "POSCAR data string"  // Crystal structure in POSCAR format
  }

• Output:

  {
    "prediction": 2.5,   // Band gap in eV
    "confidence": 98.5   // Confidence in percentage (0-100)
  }

Authentication

Include your API key in the request header:

X-API-Key: your_api_key

Error Responses

• 400 Bad Request: Invalid input format
• 401 Unauthorized: Missing or invalid API key
• 429 Too Many Requests: Rate limit exceeded
• 500 Server Error: Internal issue (contact support)

Installation (Local Development)

1. Clone the Repository:

   git clone https://github.com/DarkStarStrix/scimlhub.git
   cd scimlhub

2. Install Dependencies:

   pip install -r requirements.txt

3. Run the API Locally:

   python app/main.py

   The API will be available at http://localhost:8000.

4. Docker Deployment (Optional):

   docker-compose -f docker/docker-compose.yml up --build

Key Features

• Fast: Average response time ~50ms
• Accurate: >95% accuracy across domains
• Reliable: Confidence scores with every prediction
• Scalable: Supports millions of requests daily
• Secure: SOC2 Type II compliant

Example Outputs

Protein Structure

{
  "prediction": "H",
  "confidence": 80.56
}

Astrophysics

{
  "prediction": "GALAXY",
  "confidence": 97.29
}

Materials Science

{
  "prediction": 2.5,
  "confidence": 98.5
}

Use Cases

• Biology: Protein design, drug discovery
• Astrophysics: Stellar classification, exoplanet research
• Materials Science: Material discovery, energy applications

Pricing

Plan	Requests/Month	Price
Free	300	$0
Premium-1K	1,000	$50/month
Premium-5K	5,000	$35/month
Premium-10K	10,000	$25/month
Enterprise	Unlimited	Custom

Resources

• API Docs
• How to Use
• Changelog
• Examples

Enterprise Support

For custom models, on-premise deployment, or integration help, email: [email protected].

Support

• Email: [email protected]
• Discord: Join

License

Commercial license - see LICENSE