Octopus Inspired Optimization (OIO): A Hierarchical Framework for Navigating Protein Fitness Landscapes

Navigating the vast, rugged, and multi-modal fitness landscapes of protein sequences presents a formidable challenge for computational protein engineering, often trapping algorithms in suboptimal solutions. Existing methods struggle with the exploration-exploitation dilemma, failing to synergize global search with deep local refinement. To overcome this critical barrier, we introduce the Octopus Inspired Optimization (OIO), a novel hierarchical metaheuristic that mimics the octopus's unique neural architecture of centralized control and decentralized, parallel execution. OIO's "individual-tentacle-sucker" framework provides an intrinsic unification of global exploration and parallelized local exploitation, making it structurally ideal for complex combinatorial problems. We validated OIO's efficacy through a rigorous three-tiered experimental framework. On a real-world Green Fluorescent Protein (GFP) design benchmark, OIO surpassed a comprehensive suite of 15 competing metaheuristics, including 7 classic algorithms and 8 state-of-the-art methods from the past two years, delivering performance comparable only to a specialized local search algorithm. This success is explained by its fundamental strengths: OIO ranked first on the NK-Landscape benchmark, confirming its architectural suitability for protein-like fitness landscapes, and also ranked first on the gold-standard CEC2022 benchmark, demonstrating the raw power and efficiency of its optimization engine. OIO establishes a new, nature-inspired paradigm for protein engineering, offering a robust and powerful tool with significant potential for advancing therapeutic and enzymatic design.