Predicting Systemic Lupus Erythematosus (SLE)-Associated Epitopes using Deep Learning

Nilay Bhatt, Xincheng Cai, Tom Shin, Luning Yang

Abstract

Systemic Lupus Erythematosus (SLE) is a global health issue with diverse symptoms and serious complications. Current research is mainly focused on identifying epitopes to uncover autoimmune triggers, which contributes to the development of precision medicine for SLE. While the computational methods for epitope prediction have progressed, the complexity of SLE makes specific predictions challenging, which eventually hinders therapy development and drug discovery. Although extensive experimental data are available on certain open-source biomedical datasets, predicting SLE-associated epitopes has not been fully leveraged by traditional ML-based methods. Our project aims to leverage state-of-the-art Deep learning models like ProtBERT, ReLSO, BEETLE and MITNet to predict immunogenic potential of an epitope for SLE-related outcomes.

Navigate

/Epitope_Prediction_DL-main
├── Data
│   ├── clean_data.csv
│   ├── data.csv
│   ├── Data_processing_EDA.ipynb
├── MITNet_training
│   ├── biomitnet_adam_model.pth
│   ├── biomitnet_adamW_model.pth
│   ├── clean_data.csv
│   ├── mitnet_adamw_model.pth
│   ├── performance_ctc_base.csv
│   ├── performance_mitnet_all.csv
│   ├── test_model_bioinfo.ipynb
│   └── test_model.ipynb
├── performance_all.csv
├── ProtBert
│   ├── ProtBERT_full.ipynb
└── RELSO
    └── ReLSO (full).ipynb

Acknowledgement

This project was prepared as a part of the course CB&B 633/ CPSC 552 'Deep Learning Theory and Applications' at Yale.