This repository contains a PyTorch implementation of a multitask deep neural network (MT-DNN) model designed to predict multiple ADME (Absorption, Distribution, Metabolism, Excretion) properties of chemical compounds simultaneously from their SMILES representations.

• Introduction
• Model Architecture
• Dataset
• Installation
• Usage
• Training
• Evaluation
• Results (Optional)
• Contributing
• License

ADME properties are crucial for understanding a drug's behavior in the body and are essential in drug discovery and development. This project leverages machine learning, specifically a multitask deep neural network, to predict six ADME endpoints:

• Absorption:
  • Solubility
• Distribution:
  • Human Plasma Protein Binding (hPPB)
  • Rat Plasma Protein Binding (rPPB)
• Metabolism:
  • Human Liver Microsomes (HLM)
  • Rat Liver Microsomes (RLM)
• Excretion/Transport:
  • MDR1-MDCK ER (Multidrug Resistance Protein 1 Efflux Ratio)

The MT-DNN model consists of the following components:

• SMILES Encoding: Input SMILES strings are encoded into integer sequences.
• Embedding Layer: Learns a dense representation for each character in the SMILES vocabulary.
• Shared LSTM Layers: Two LSTM layers capture sequential dependencies in the SMILES strings to learn a shared representation of the molecule.
• Task-Specific Heads: Six linear layers, each responsible for predicting one of the ADME endpoints.

• Simulated ADME Data: The project includes a script for generating synthetic ADME data with SMILES representations and corresponding ADME values.
• Real-World Data: You can replace the simulated data with real-world datasets like Tox21, PubChem, or custom datasets from pharmaceutical companies.

1. Clone the repository: git clone [repository url]
2. Install dependencies: pip install -r requirements.txt

1. Data Preparation:
   • Use prepare_data.py to preprocess and split your dataset into training and validation sets.
2. Model Training:
   • Run train.py to train the MT-DNN model. Adjust hyperparameters and training options as needed.
3. Evaluation:
   • Use evaluate.py to evaluate the trained model on the test set and compute metrics like MSE and R^2.

Refer to the comments in the train.py script for detailed instructions on training the model.

Refer to the comments in the evaluate.py script for instructions on evaluating the model.

Add a section to summarize your findings, including:

• Model performance on the test set (e.g., MSE, R^2 for each task).
• Any insights gained from feature importance analysis or model interpretation.
• Comparison with other baseline models (if applicable).

Contributions are welcome! Feel free to open issues or submit pull requests.

This project is licensed under the [LICENSE NAME] License.