🧬 MOLE-AI

AI Platform for Molecular Property Prediction and Computational Drug Discovery

Accelerating AI-driven drug discovery through machine learning, cheminformatics, and molecular intelligence.

Overview

MOLE-AI is an open-source platform for AI-powered molecular property prediction and computational drug discovery.

The project combines cheminformatics, machine learning, deep learning, and computational chemistry into a unified framework for developing molecular prediction pipelines, QSAR models, molecular similarity analysis, and drug discovery workflows.

The primary goal of MOLE-AI is to provide researchers, students, and developers with an extensible toolkit for molecular analysis and predictive modeling while promoting reproducible and open scientific research.

🌐 Interactive Web Application

MOLE-AI includes a Streamlit-based interface for molecular analysis and AI prediction.

🚀 Live Demo

**Try MOLE-AI online:**https://mole-ai-toolkit.streamlit.app/

Current capabilities:

✅ SMILES-based molecular analysis
✅ Molecular fingerprint generation
✅ QSAR-based pIC50 prediction
✅ Molecular structure visualization
✅ Prediction history export

Launch locally:

streamlit run mole_ai/app.py

Key Highlights

🧬 Molecular preprocessing using RDKit
🧪 SMILES validation and molecular analysis
📊 Molecular descriptor calculation
🧠 Machine Learning (QSAR)
🤖 Deep Learning models
🕸️ Graph Neural Network foundation
🔬 Transformer-based molecular models
💊 ADMET prediction utilities
⚗️ Molecular optimization
🧩 Molecular similarity search
📈 Prediction reporting
🌐 Interactive Streamlit web interface
🧪 Automated testing with pytest
📚 Modular Python API
🚀 Open-source and extensible architecture

Current Release

MOLE-AI v1.0.0

The first stable release includes a complete molecular processing pipeline covering:

Molecular preprocessing
Descriptor generation
Fingerprint generation
Feature engineering
ChEMBL data processing
QSAR model training
Prediction pipeline
Model evaluation
Batch prediction
Prediction reports
Dataset management
Molecular similarity search
Experiment tracking
Model registry
Streamlit interface
Documentation
Automated testing

🚀 Features

MOLE-AI provides a modular collection of tools for molecular analysis, machine learning, and AI-assisted drug discovery.

🖥️ Streamlit Application Features

MOLE-AI includes an interactive Streamlit web application for molecular analysis and AI-based molecular property prediction.

The interface is organized into four main modules:

🧪 Tab 1 — Molecular Analysis

Analyze molecules directly from SMILES input.

Features:

SMILES validation using RDKit
Molecular structure visualization
Molecular descriptors calculation
Drug-likeness properties
Lipinski rule analysis

Screenshot:

🧬 Tab 2 — Fingerprint Generation

The Fingerprint Generation module converts molecules into machine-learning-ready molecular representations.

Features:

✅ Morgan fingerprint generation
✅ 2048-bit molecular fingerprints
✅ Molecular feature encoding
✅ Visualization of active fingerprint bits

Screenshot:

🤖 Tab 3 — AI Molecular Prediction

The AI Prediction module uses a QSAR machine learning model to estimate molecular activity.

Current prediction workflow:

  text
SMILES Input
      ↓
RDKit Molecular Processing
      ↓
Morgan Fingerprint Generation
      ↓
Random Forest QSAR Model
      ↓
Predicted pIC50 Activity

Features:

✅ Random Forest regression model ✅ Morgan fingerprint-based prediction ✅ pIC50 activity prediction ✅ Activity interpretation ✅ Prediction history tracking ✅ CSV export capability

Screenshot

📚 Tab 4 — About MOLE-AI

Provides an overview of the MOLE-AI platform, including its scientific workflow, technology stack, and future development roadmap.

Features:

🧬 AI-driven drug discovery workflow
🧪 Cheminformatics and machine learning pipeline overview
🛠 Technology stack information
📊 Project architecture overview
🗺 Future development roadmap

The About section summarizes how MOLE-AI integrates:

Molecular Input ↓ RDKit Processing ↓ Molecular Features ↓ Machine Learning Models ↓ Property Prediction ↓ Drug Discovery Insights

Screenshot

🧬 Cheminformatics

Feature	Description	Status
✅ SMILES Validation	Validate molecular SMILES strings using RDKit	✔
✅ Molecular Parsing	Convert SMILES into RDKit molecule objects	✔
✅ Descriptor Calculation	Generate physicochemical descriptors	✔
✅ Morgan Fingerprints	Circular fingerprints for molecular representation	✔
✅ Molecular Similarity	Tanimoto similarity search	✔
✅ Molecular Visualization	Interactive molecular structure display	✔

🤖 Machine Learning

Feature	Description	Status
✅ Feature Engineering	Molecular feature generation pipeline	✔
✅ QSAR Modelling	Random Forest regression models	✔
✅ Model Training	Train predictive molecular models	✔
✅ Batch Prediction	Predict multiple molecules simultaneously	✔
✅ Model Evaluation	MAE, RMSE and R² metrics	✔
✅ Prediction Reports	Automatic prediction summaries	✔
✅ Model Registry	Save and manage trained models	✔

🤖 QSAR Model

Current prediction engine:

Component	Description
Algorithm	Random Forest Regression
Input Features	2048-bit Morgan Fingerprints
Prediction	pIC50 Activity
Chemistry Engine	RDKit
Model Type	Supervised Machine Learning

Workflow:

SMILES

↓

RDKit Molecular Processing

↓

Morgan Fingerprint

↓

Random Forest Model

↓

Predicted pIC50

🧠 Deep Learning

Feature	Description	Status
✅ Feed-forward Neural Networks	Deep QSAR models	✔
✅ Graph Neural Network Foundation	Molecular graph learning modules	✔
✅ Transformer Models	Transformer-based molecular prediction	✔
🔄 Large Language Models	Future integration	Planned

💊 Drug Discovery

Feature	Description	Status
✅ ADMET Prediction	Drug-likeness utilities	✔
✅ Molecular Optimization	Lead optimization tools	✔
✅ Molecular Generation	AI-assisted molecule generation	✔
✅ Docking Utilities	Docking workflow support	✔
✅ Candidate Ranking	Rank compounds by predicted performance	✔
✅ ChEMBL Processing	Dataset preprocessing pipeline	✔

🌐 User Interface

Feature	Description	Status
✅ Streamlit Application	Interactive web interface	✔
✅ Python API	Easy integration into Python projects	✔
✅ Command-Line Interface	Terminal-based workflows	✔
✅ Modular Package Design	Reusable Python modules	✔

📊 Current Capabilities

MOLE-AI currently supports the complete workflow below:

SMILES Input
      │
      ▼
Validation
      │
      ▼
Descriptor Generation
      │
      ▼
Fingerprint Generation
      │
      ▼
Feature Engineering
      │
      ▼
Machine Learning Prediction
      │
      ▼
Model Evaluation
      │
      ▼
Prediction Report

⭐ Why MOLE-AI?

MOLE-AI was designed with the following principles:

🧬 Modular architecture
⚡ Fast molecular preprocessing
🤖 AI-first drug discovery workflows
🔬 Reproducible computational research
📚 Open-source development
🧪 Extensible machine learning pipelines
🌐 Interactive web interface
📦 Easy integration into existing research projects

📈 Project Statistics

Metric	Value
Programming Language	Python 3.11+
Core Cheminformatics Library	RDKit
Machine Learning Framework	Scikit-learn
Deep Learning Framework	PyTorch
Web Framework	Streamlit
Documentation	Markdown
Testing Framework	Pytest
License	MIT
Current Release	v1.0.0

🏗️ System Architecture

MOLE-AI follows a modular architecture that separates molecular preprocessing, feature engineering, machine learning, deep learning, and drug discovery utilities into reusable components.

The platform is designed to be scalable, allowing researchers to extend individual modules without affecting the rest of the system.

🧬 Overall Architecture

                          ┌───────────────────────┐
                          │       User/API        │
                          └──────────┬────────────┘
                                     │
                     ┌───────────────┴───────────────┐
                     │                               │
              Streamlit App                   Command Line Interface
                     │                               │
                     └───────────────┬───────────────┘
                                     │
                           Python API Layer
                                     │
        ┌────────────────────────────┼────────────────────────────┐
        │                            │                            │
        ▼                            ▼                            ▼
 Chemistry Module            Feature Engineering          Machine Learning
        │                            │                            │
        ▼                            ▼                            ▼
 Descriptors                  Molecular Features          QSAR Models
 Fingerprints                 Dataset Processing          Model Training
 SMILES Validation            ChEMBL Pipeline             Prediction
        │                            │                            │
        └────────────────────────────┼────────────────────────────┘
                                     │
                                     ▼
                           Drug Discovery Modules
                                     │
        ┌───────────────┬───────────────┬───────────────┐
        ▼               ▼               ▼               ▼
     ADMET       Similarity Search  Docking      Optimization
                                     │
                                     ▼
                           Prediction Reports
                                     │
                                     ▼
                              Final Results

📂 Project Modules

MOLE-AI is organized into independent modules.

Module	Purpose
chem	Molecular preprocessing and cheminformatics
features	Feature engineering and descriptor pipelines
models	Machine learning and deep learning models
data	Dataset loading and processing
examples	Example workflows
tests	Automated unit testing
docs	Documentation
app.py	Streamlit web application
cli.py	Command-line interface

🔬 Drug Discovery Pipeline

Input SMILES
      │
      ▼
SMILES Validation
      │
      ▼
RDKit Molecule
      │
      ▼
Descriptor Generation
      │
      ▼
Morgan Fingerprints
      │
      ▼
Feature Engineering
      │
      ▼
Machine Learning Models
      │
      ▼
Deep Learning Models
      │
      ▼
Property Prediction
      │
      ▼
ADMET Prediction
      │
      ▼
Similarity Search
      │
      ▼
Docking Utilities
      │
      ▼
Lead Optimization
      │
      ▼
Prediction Report

🔄 Data Flow

The figure below summarizes the complete data flow implemented in MOLE-AI.

SMILES
   │
   ▼
RDKit
   │
   ▼
Descriptors
   │
   ▼
Fingerprints
   │
   ▼
Feature Matrix
   │
   ▼
Machine Learning
   │
   ▼
Prediction
   │
   ▼
Evaluation
   │
   ▼
Visualization

📦 Software Design Principles

MOLE-AI has been developed following modern software engineering practices.

✅ Modular architecture
✅ Reusable components
✅ Object-oriented design
✅ Python package structure
✅ Automated testing
✅ Documentation-first development
✅ Scalable project organization
✅ Easy integration with external tools

🎯 Design Goals

The architecture was designed to support:

AI-assisted drug discovery
Molecular property prediction
Reproducible computational workflows
Educational bioinformatics projects
Future deep learning extensions
Open-source community contributions

⚙️ Installation

System Requirements

MOLE-AI has been tested with the following environment.

Requirement	Version
Python	3.11+
Conda	Latest
RDKit	Latest
Streamlit	Latest
Scikit-learn	Latest
PyTorch	Latest

📥 Clone Repository

git clone https://github.com/Mehwish55/MOLE-AI.git

cd MOLE-AI

🐍 Create Conda Environment

conda create -n mole-ai python=3.11

conda activate mole-ai

📦 Install Dependencies

pip install -r requirements.txt

✅ Verify Installation

Run:

pytest -v

If all tests pass successfully, MOLE-AI has been installed correctly.

🚀 Quick Start

The simplest way to use MOLE-AI is through the Python API.

from mole_ai.chem.smiles import validate_smiles

print(validate_smiles("CCO"))

Output

True

🧬 Python API

Validate a SMILES string

from mole_ai.chem.smiles import validate_smiles

validate_smiles("CCO")

Generate molecular descriptors

from mole_ai.chem.smiles import smiles_to_mol
from mole_ai.chem.descriptors import calculate_descriptors

mol = smiles_to_mol("CCO")
calculate_descriptors(mol)

Generate Morgan fingerprints

from mole_ai.chem.smiles import smiles_to_mol
from mole_ai.chem.fingerprints import generate_morgan_fingerprint

mol = smiles_to_mol("CCO")
generate_morgan_fingerprint(mol)

Predict molecular properties

from mole_ai.models.predict import predict_from_smiles

prediction = predict_from_smiles("CCO")
print(prediction)

💻 Command-Line Interface

Display the installed version.

python -m mole_ai.cli --version

Run molecular prediction.

python -m mole_ai.cli \
    --predict \
    --model models/model.pkl \
    --input data/features.csv \
    --output predictions.csv

🌐 Streamlit Web Interface

MOLE-AI includes an interactive Streamlit application for molecular property prediction.

Launch the application

streamlit run mole_ai/app.py

Open your browser

http://localhost:8501

🖥️ Web Interface Features

The Streamlit application currently supports:

✅ SMILES input
✅ Molecular validation
✅ Property prediction
✅ Prediction summary
✅ Interactive interface

📸 Web Interface Preview

Screenshot coming soon

After launching the application you can enter a molecular SMILES string such as:

CCO

The application validates the molecule and returns prediction results through an interactive interface.

📁 Example Files

The repository includes example scripts demonstrating common workflows.

examples/

build_features.py

prediction_example.py

training_example.py

🔬 Typical Workflow

Clone Repository

↓

Install Dependencies

↓

Prepare Dataset

↓

Generate Features

↓

Train Model

↓

Evaluate Model

↓

Predict Properties

↓

Generate Reports

↓

Visualize Results

⚡ Performance

Current implementation supports:

Molecular preprocessing
Descriptor generation
Fingerprint calculation
QSAR prediction
Batch prediction
Prediction reports
Molecular similarity
Interactive web interface

The modular design allows future expansion with additional machine learning models and drug discovery workflows.

📂 Project Structure

The repository follows a modular architecture designed for scalability, maintainability, and reproducible computational research.

MOLE-AI/
│
├── assets/
│   └── images/
│       ├── logo.png
│       └── github-banner.png
│
├── data/
│   ├── raw/
│   ├── processed/
│   └── external/
│
├── docs/
│   ├── index.md
│   ├── api.md
│   ├── user_guide.md
│   └── workflow.md
│
├── examples/
│   ├── build_features.py
│   ├── prediction_example.py
│   └── training_example.py
│
├── mole_ai/
│   ├── chem/
│   │   ├── descriptors.py
│   │   ├── fingerprints.py
│   │   └── smiles.py
│   │
│   ├── data/
│   │
│   ├── features/
│   │
│   ├── models/
│   │   ├── train.py
│   │   ├── predict.py
│   │   ├── evaluate.py
│   │   ├── registry.py
│   │   ├── tuning.py
│   │   ├── explainability.py
│   │   ├── graph_neural_network.py
│   │   ├── transformer.py
│   │   ├── admet.py
│   │   ├── optimization.py
│   │   └── docking.py
│   │
│   ├── app.py
│   └── cli.py
│
├── notebooks/
│
├── tests/
│
├── requirements.txt
├── pyproject.toml
├── LICENSE
└── README.md

📦 Package Organization

Directory	Description
assets/	Images, logos, and project branding
data/	Raw, processed, and external datasets
docs/	Project documentation
examples/	Example scripts and tutorials
mole_ai/	Core Python package
notebooks/	Jupyter notebooks
tests/	Automated unit tests
requirements.txt	Python dependencies
pyproject.toml	Project configuration
README.md	Project overview

🧩 Core Modules

🧬 Chemistry Module

Responsible for molecular preprocessing.

Functions include:

SMILES validation
RDKit molecule conversion
Molecular descriptors
Morgan fingerprints
Similarity calculations

📊 Feature Engineering

Responsible for preparing machine learning inputs.

Includes:

Descriptor extraction
Fingerprint generation
Feature matrices
Dataset preparation
ChEMBL preprocessing

🤖 Machine Learning

Implements predictive models.

Includes:

Random Forest
Model training
Model evaluation
Batch prediction
Prediction reports
Hyperparameter tuning

🧠 Deep Learning

Provides neural-network-based models.

Includes:

Feed-forward neural networks
Graph neural network foundation
Transformer models

💊 Drug Discovery

Utilities supporting molecular design.

Includes:

ADMET prediction
Docking utilities
Molecular optimization
Candidate ranking
Similarity search

🌐 User Interfaces

Two interfaces are available.

Python API

Designed for researchers and developers.

Streamlit Application

Designed for interactive molecular prediction.

📊 Development Statistics

Category	Status
Python Package	✅
RDKit Integration	✅
Machine Learning	✅
Deep Learning	✅
Documentation	✅
Unit Testing	✅
GitHub Release	✅
Streamlit Interface	✅
CLI Support	✅

🧪 Testing

The project uses pytest for automated testing.

Current coverage includes:

Molecular preprocessing
Descriptor calculation
Fingerprint generation
Feature engineering
Model training
Prediction
Evaluation
Reporting
Dataset utilities
Similarity search

Run all tests:

pytest -v

Run an individual test:

pytest tests/test_predict.py -v

📚 Documentation

Project documentation is available in the docs/ directory.

docs/

├── index.md
├── api.md
├── workflow.md
└── user_guide.md

Documentation includes:

Installation Guide
User Guide
API Reference
Workflow Documentation

📈 Development Philosophy

MOLE-AI has been developed with the following goals:

Modular architecture
Reusable Python components
Reproducible computational workflows
Open-source collaboration
Extensible AI framework
Educational usability
Production-ready project organization

🛠️ Technologies

MOLE-AI is built using a modern scientific Python ecosystem for computational chemistry, machine learning, and bioinformatics.

Programming Language

Python 3.11+

Cheminformatics

RDKit

Machine Learning

Scikit-learn
NumPy
Pandas

Deep Learning

PyTorch
PyTorch Geometric (foundation for future graph-based models)

Data Processing

ChEMBL
PubChem
Protein Data Bank (PDB)

Visualization

Streamlit
Matplotlib

Development

Git
GitHub
Pytest
Conda

📚 Datasets

MOLE-AI is designed to work with publicly available molecular datasets.

Supported datasets include:

ChEMBL
PubChem
Protein Data Bank (PDB)

Typical applications include:

QSAR modelling
Molecular property prediction
Feature engineering
Molecular similarity analysis
Drug discovery research

🧪 Quality Assurance

The project follows software engineering best practices.

✅ Modular architecture
✅ Version control using Git
✅ Automated unit testing
✅ Reproducible workflows
✅ Open-source development
✅ Continuous documentation

🗺️ Roadmap

The following features are planned for future releases.

Version 1.1

Enhanced Streamlit dashboard
Molecular visualization
Improved prediction reports
Additional QSAR algorithms

Version 1.2

Advanced Graph Neural Networks
Transformer fine-tuning
Additional molecular descriptors
Model comparison dashboard

Version 2.0

Large molecular foundation models
AI-assisted molecular generation
Multi-objective optimization
Cloud deployment
REST API
Docker support
Interactive molecular visualization
Automated benchmarking

🤝 Contributing

Contributions are welcome.

If you would like to contribute:

Fork the repository
Clone your fork

git clone https://github.com/<your-username>/MOLE-AI.git

Create a feature branch

git checkout -b feature/new-feature

Commit your changes

git commit -m "Add new feature"

Push your branch

git push origin feature/new-feature

Open a Pull Request

Please ensure that:

Code follows project style
Tests pass successfully
Documentation is updated where appropriate

📖 Citation

If you use MOLE-AI in your research, teaching, or publications, please cite this repository.

Shafiq, M.

MOLE-AI:
An Open-Source Platform for Molecular Property Prediction
and Computational Drug Discovery.

GitHub Repository.

https://github.com/Mehwish55/MOLE-AI

A formal citation (e.g., DOI via Zenodo) may be added in future releases.

📄 License

This project is licensed under the MIT License.

See the LICENSE file for additional information.

🙋 About the Author

Mehwish Shafiq

Computational Biologist | Bioinformatics Researcher | AI for Drug Discovery

Research interests include:

Computational Drug Discovery
Bioinformatics
Machine Learning
Artificial Intelligence
Molecular Property Prediction
Multi-Objective Drug Design
Computational Biology

📬 Contact

For questions, feature requests, or collaboration opportunities:

Open a GitHub Issue
Submit a Pull Request
Connect through GitHub

Repository:

https://github.com/Mehwish55/MOLE-AI

⭐ Support the Project

If you find MOLE-AI useful:

⭐ Star the repository
🍴 Fork the project
🧪 Try the examples
📝 Report bugs
💡 Suggest new features
🤝 Contribute improvements

Your support helps improve the project and encourages future development.

🎯 Project Vision

MOLE-AI aims to provide an open, modular, and extensible platform that bridges computational chemistry, bioinformatics, and artificial intelligence.

The long-term vision is to support researchers, students, and developers in accelerating molecular discovery through reproducible, AI-driven workflows.

🧬 MOLE-AI

Accelerating AI-driven Drug Discovery

Made with ❤️ using Python, RDKit, Machine Learning, and Open Science.

Name		Name	Last commit message	Last commit date
.devcontainer		.devcontainer
assets/images		assets/images
data		data
docs		docs
examples		examples
mole_ai		mole_ai
scripts		scripts
tests		tests
.dockerignore		.dockerignore
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CITATION.cff		CITATION.cff
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
LICENSE		LICENSE
README.md		README.md
README.md.save		README.md.save
README_v2.md		README_v2.md
RELEASE_NOTES.md		RELEASE_NOTES.md
SPACE_README.md		SPACE_README.md
app.py		app.py
environment.conda.yml		environment.conda.yml
pyproject.toml		pyproject.toml
render.yaml		render.yaml
requirements.txt		requirements.txt
runtime.txt		runtime.txt
v2_app.py		v2_app.py

Latest commit

History

Folders and files

Repository files navigation

🧬 MOLE-AI

AI Platform for Molecular Property Prediction and Computational Drug Discovery

Overview

🌐 Interactive Web Application

🚀 Live Demo

Key Highlights

Current Release

MOLE-AI v1.0.0

🚀 Features

🖥️ Streamlit Application Features

🧪 Tab 1 — Molecular Analysis

🧬 Tab 2 — Fingerprint Generation

🤖 Tab 3 — AI Molecular Prediction

📚 Tab 4 — About MOLE-AI

🧬 Cheminformatics

🤖 Machine Learning

🤖 QSAR Model

🧠 Deep Learning

💊 Drug Discovery

🌐 User Interface

📊 Current Capabilities

⭐ Why MOLE-AI?

📈 Project Statistics

🏗️ System Architecture

🧬 Overall Architecture

📂 Project Modules

🔬 Drug Discovery Pipeline

🔄 Data Flow

📦 Software Design Principles

🎯 Design Goals

⚙️ Installation

System Requirements

📥 Clone Repository

🐍 Create Conda Environment

📦 Install Dependencies

✅ Verify Installation

🚀 Quick Start

🧬 Python API

💻 Command-Line Interface

🌐 Streamlit Web Interface

🖥️ Web Interface Features

📸 Web Interface Preview

📁 Example Files

🔬 Typical Workflow

⚡ Performance

📂 Project Structure

📦 Package Organization

🧩 Core Modules

🧬 Chemistry Module

📊 Feature Engineering

🤖 Machine Learning

🧠 Deep Learning

💊 Drug Discovery

🌐 User Interfaces

Python API

Streamlit Application

📊 Development Statistics

🧪 Testing

📚 Documentation

📈 Development Philosophy

🛠️ Technologies

Programming Language

Cheminformatics

Machine Learning

Deep Learning

Data Processing

Visualization

Development

📚 Datasets

🧪 Quality Assurance

🗺️ Roadmap

Version 1.1

Version 1.2

Version 2.0

🤝 Contributing

📖 Citation