🧬 Antibiotic Resistance Prediction

This project explores machine learning methods to predict antibiotic resistance classes from protein sequences.
It uses protein-level encoding (amino acid one-hot representation) and deep learning models (LSTM/GRU/Transformer) to classify sequences into known antibiotic resistance categories.

---

Project Structure

AntibioticResistance/
├── lib/
│   ├── encoding.py        # Protein one-hot encoder & data preparation
│   ├── model.py           # Deep learning model (LSTM/GRU/Transformer)
├── main.ipynb             # Jupyter notebook for experiments
├── requirements.txt       # Dependencies
└── README.md              # Project description

---

Installation

Clone the repository and install dependencies:

git clone https://github.com/YOUR_USERNAME/AntibioticResistance.git
cd AntibioticResistance
pip install -r requirements.txt

---

Data

The input dataset should contain protein sequences with metadata. Example format:

Allele	Gene family	Product name	Class	Sequence protein
aac(3)-VIIIa	aac(3)-VIII	aminoglycoside N-acetyltransferase AAC(3)-VIIIa	AMINOGLYCOSIDE	MDEKELIERAGG...
aac(6’)-32	aac(6’)	aminoglycoside N-acetyltransferase AAC(6’)-32	AMINOGLYCOSIDE	MSPSKTPVTLR...
qnrS1	qnr	quinolone resistance protein QnrS1	QUINOLONE	MTQDLMTLFNV...

---

Pipeline

1. Encoding

   • Protein sequences → one-hot vectors (20 amino acids)
   • Padding/truncation to fixed length

2. Model

   • Sequence classification with LSTM/GRU/Transformer
   • Multi-class classification over antibiotic resistance Class

3. Training

   from lib.encoding import prepare_data
   from lib.model import ResistanceModel
   
   X, y, label_encoder = prepare_data(df, max_len=500)
   model = ResistanceModel(max_len=500, num_classes=len(label_encoder.classes_))
   history = model.train(X, y, epochs=10, batch_size=32, validation_split=0.2)

4. Prediction

   preds = model.predict(X[:5])
   label_encoder.inverse_transform(preds.argmax(axis=1))

---

Saving & Loading Models

Save:

model.model.save("./model/resistance_model.h5")

Load:

from tensorflow.keras.models import load_model
loaded_model = load_model("./model/resistance_model.h5")

---

🧪Example Output

Predictions: ['AMINOGLYCOSIDE', 'QUINOLONE', 'LIPOPEPTIDE']

---

TODO

• Improve dataset preprocessing
• Try Transformer-based encoders
• Evaluate with larger benchmark datasets
• Deploy as an API for resistance prediction

---

License

MIT License