IFNg_DeepKG: A Novel Model for Identifying Interferon-Gamma-Inducing Epitopes Using Knowledge Graph RAG in Biomedical...

Area of Science:

• Computational immunology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Accurate identification of interferon-gamma-inducing epitopes (IFNgIE) is crucial for designing effective vaccines and immunotherapies.
• Current computational models often overlook essential biological context, limiting their predictive accuracy for epitope immunogenicity.

Purpose of the Study:

• To develop a novel deep learning framework, IFNg_DeepKG, that integrates sequence-based patterns with rich biological context for improved IFNgIE prediction.
• To enhance the design of next-generation vaccines and immunotherapies through more accurate epitope identification.

Main Methods:

• IFNg_DeepKG framework combines a pretrained protein language model (ESM2), a custom knowledge graph (KG) with Retrieval-Augmented Generation (RAG), and a multiscale convolutional neural network (MSCNN).
• The RAG-KG enriches sequence embeddings with external biological information (protein of origin, host, disease association) to improve immunogenicity predictions.

Main Results:

• IFNg_DeepKG achieved superior performance on independent test datasets, with AUCs of 0.99 (Human H_IFNgInd1) and 0.95 (Mouse M_IFNgInd1).
• Demonstrated strong cross-species generalization with AUCs of 0.94 (H_IFNgInd2) and 0.93 (M_IFNgInd2) on challenging datasets.
• Successfully identified and classified clinically relevant epitopes associated with diseases like COVID-19 and Alzheimer's disease.

Conclusions:

• IFNg_DeepKG significantly advances computational immunology by bridging sequence-based features and biological contexts.
• The framework offers a scalable and powerful platform for rational epitope discovery, precision medicine, and the development of novel vaccines and immunotherapies.