SiaScoreNet

A Siamese neural network-based model integrating prediction scores for HLA-peptide interaction prediction

Overview

SiaScoreNet is a deep learning model that predicts the binding affinity between HLA class I proteins and peptides. It integrates sequence embeddings and ensemble scores from existing predictors, using a Siamese-like architecture.

Data availability

All datasets used for training, validation, and testing of the SiaScoreNet model are publicly available on Zenodo Repository: https://doi.org/10.5281/zenodo.15606313

---

How to Run

Follow these steps to perform binding prediction using SiaScoreNet:

1. Install Requirements

pip install -r requirements.txt

2. Prepare Input File

Create a CSV file named test_example.csv with two columns:

• peptide
• HLA

Example:

peptide, HLA

LLFGYPVYV, HLA-A*02:01

NLVPMVATV, HLA-B*07:02

3. Run Feature Extraction

Extract embeddings and ensemble scores by running:

python feature_extraction.py --input test_example.csv --output features_extracted.csv

This will generate the file features_extracted.csv, which contains:

• 9 IEDB scores for ensemble
• ESM embeddings of the peptide
• ESM embeddings of the HLA sequence

4. Run Prediction

Use the extracted features to make predictions:

python predict.py --input features_extracted.csv --output predictions.csv

The result will be saved in predictions.csv, containing:

peptide,HLA,predicted_score,predicted_label

Where:

• predicted_score is the probability predicted by the model.
• predicted_label is the binary class (0 or 1) based on threshold 0.5.