Quantum Optics Seminar: Timothy P. Jenkins, DTU

Unlocking Peptide Design with Quantum-Enhanced Generative AI:

Charting the Future of Drug Discovery and Vaccine Design

 Recent advancements in generative machine learning have accelerated the design of therapeutics, reducing development timelines from months to weeks. However, quantum computing, with its unique capability to sample from high-dimensional distributions beyond classical reach, has the potential to further enhance these models by introducing complex, non-classical correlations into the learning process.

We have produced the first experimental implementation of a quantum-classical algorithm designed to generate MHC-binding peptides for vaccine applications. By integrating a boson sampling quantum processor with a generative adversarial network (GAN), our approach creates a hybrid model that leverages quantum-enhanced distributions as latent spaces, resulting in improved learning dynamics and superior model performance.

This quantum-assisted GAN framework outperforms classical GANs of similar architecture, generating a diverse range of MHC-binding peptides with potential as immune-modulating agents. Our results underscore the promise of quantum-enhanced generative models, showcasing how quantum computing can advance peptide design by optimizing learning efficiency and sequence diversity, with direct implications for the biotechnology and pharmaceutical fields.