APCyc: Property-Informed Design of Cyclic Peptides via Automated Cyclization

APCyc is a newly introduced framework for the de novo design of cyclic peptides, addressing the challenges of identifying optimal cyclization patterns and linkage sites while optimizing multiple physicochemical properties. It utilizes an expanded residue vocabulary and encodes cyclization-specific information, enabling the model to learn target-dependent cyclization preferences and perform effective multi-property optimization. This development is significant for practitioners as it enhances the design of therapeutic cyclic peptides, potentially improving drug stability and binding affinity in drug discovery applications.