epiGPTope: A Machine Learning-Based Epitope Generator and Classifier.

Area of Science:

• Immunoinformatics
• Computational Biology
• Artificial Intelligence in Biotechnology

Background:

• Epitopes are critical for developing immunotherapies, vaccines, and diagnostics.
• Designing synthetic epitope libraries is hindered by the vast sequence space, making experimental screening infeasible.

Purpose of the Study:

• To present epiGPTope, a large language model for generating novel linear epitope sequences.
• To develop a computational approach for accelerating epitope discovery and library design.

Main Methods:

• Fine-tuning a large language model (epiGPTope) on linear epitope data.
• Utilizing generative models to produce novel epitope-like sequences with analogous statistical properties.
• Training statistical classifiers to predict the origin (bacterial or viral) of epitope sequences.

Main Results:

• EpiGPTope successfully generates novel epitope-like sequences with statistical properties similar to known epitopes.
• The generative approach enables the creation of candidate epitope libraries.
• Predictive models can differentiate between bacterial and viral epitope origins, refining candidate selection.

Conclusions:

• The combination of generative and predictive models offers a powerful tool for epitope discovery.
• This AI-driven method bypasses the need for complex structural data or handcrafted features.
• The approach promises faster, more cost-effective generation and screening of synthetic epitopes for biotechnological applications.