Combining Bayesian optimization with sequence- or structure-based strategies for optimization of peptide-binding protein

This study introduces a novel Bayesian Optimization (BO) method, designed to support the design and optimization of bioactive peptides in the context od a fully automated closed-loop Design-Make-Test (DMT) pipeline. We demonstrate the capacity of this approach using the Major Histocompatibility Complex (MHC) class I receptor system as a benchmark dataset, starting with a single peptide-lead sequence in the μM IC50 range and efficiently optimizing it to approach optimal binding affinity within just 4-5 DMT cycles. We extensively evaluated its performance, varying conditions and parameters. Different sequence- and structure-based initialization strategies were also tested, to generate the initial peptide population. The developed approach can effectively handle various peptide lengths simultaneously, and also in small batches provide a valuable foundation for peptide optimization in closed-loop DMT environments. Our studies underline the potential of our BO method to efficiently navigate vast peptide sequence spaces, significantly advancing the development of new bioactive peptides. The source code of our method, Mobius, is publicly available under the Apache license at https://git.scicore.unibas.ch/schwede/mobius.