This research article presents a novel approach to peptide design by integrating protein structure prediction with Bayesian Optimization (BO). The study aims to identify peptides with unique biological properties, addressing the challenges of traditional and computational methods in peptide design. By leveraging deep learning models and BO, the method enhances the optimization of peptide sequences, guided by their predicted 3D structures. The authors use ColabFold for rapid protein modeling and a multi-objective Gaussian surrogate process to refine peptide sequences. The methodology includes a detailed algorithm that iteratively optimizes peptide sequences based on objective functions such as Solvent Accessible Surface Area (SASA), stability, and binding site ratio. The results show significant improvements in the structural similarity and biological function of optimized peptides compared to native sequences. The study concludes that this integrated approach holds promise for peptide design, though it also acknowledges limitations such as computational time and the need for further improvements in embedding techniques.This research article presents a novel approach to peptide design by integrating protein structure prediction with Bayesian Optimization (BO). The study aims to identify peptides with unique biological properties, addressing the challenges of traditional and computational methods in peptide design. By leveraging deep learning models and BO, the method enhances the optimization of peptide sequences, guided by their predicted 3D structures. The authors use ColabFold for rapid protein modeling and a multi-objective Gaussian surrogate process to refine peptide sequences. The methodology includes a detailed algorithm that iteratively optimizes peptide sequences based on objective functions such as Solvent Accessible Surface Area (SASA), stability, and binding site ratio. The results show significant improvements in the structural similarity and biological function of optimized peptides compared to native sequences. The study concludes that this integrated approach holds promise for peptide design, though it also acknowledges limitations such as computational time and the need for further improvements in embedding techniques.