Functional protein design and enhancement with ontology reinforcement iteration

2026-03-19 · Source: Machine learning : nature.com subject feeds · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Life Sciences & Biology · Depth: Expert, medium

Summary

ORI (Ontology Reinforcement Iteration) is a scalable framework designed to address the gap between computational modeling and experimental performance in protein engineering. It integrates ontology-conditioned decoding with reinforcement learning from experimental feedback (RLWF), using structured ontologies as semantic prompts to apply multi-level constraints for controllable and interpretable protein generation. ORI employs a closed-loop iterative workflow involving generation, experimental measurement, and model updating for continuous optimization under real-world objectives. The framework has demonstrated its utility in various tasks, including optimizing enzymatic activity, enhancing thermal stability, and engineering multifunctional proteins. For example, ORI engineered a lysozyme with 100-fold higher activity, a chitinase stable at 85 °C, and dual-function enzymes exhibiting both lysozyme and chitinase activities, significantly improving upon natural baselines. The source code and model weights are available on GitHub and Zenodo.

Key takeaway

For AI Researchers and Research Scientists working on protein engineering, ORI offers a robust platform to bridge the gap between computational design and real-world experimental performance. You should consider integrating ontology-conditioned decoding and reinforcement learning from experimental feedback into your protein design workflows to achieve more controllable, interpretable, and functionally superior protein variants, especially for multi-objective optimization tasks. This approach can lead to substantial improvements over natural baselines.

Key insights

ORI integrates ontology-conditioned decoding with reinforcement learning for controllable, interpretable, and experimentally validated protein design.

Principles

• Integrate semantic prompts for constrained generation.
• Utilize experimental feedback for continuous optimization.
• Employ iterative closed-loop workflows.

Method

ORI's method involves ontology-conditioned decoding to generate protein variants, followed by experimental measurement of their performance, and subsequent model updating via reinforcement learning from experimental feedback (RLWF) in a closed-loop iteration.

In practice

• Optimize enzymatic activity (e.g., lysozyme).
• Enhance protein thermal stability (e.g., chitinase).
• Engineer multifunctional proteins.

Topics

• Protein Engineering
• Reinforcement Learning
• Ontology-conditioned Decoding
• Functional Protein Design
• Multi-objective Optimization

Code references

• TencentAI4S/ori

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• GPT-5 lowers the cost of cell-free protein synthesis — GPT-5 connected to a robotic lab significantly reduced cell-free protein synthesis costs through autonomous closed-loop optimization.
• Profluent and Lilly: the next gene editor will be designed by AI — AI-designed recombinases offer a credible path to kilobase-scale DNA editing for complex genetic diseases.
• 🔬ESM: The Bitter Lesson is Coming for Proteins - Alex Rives, BioHub — Scaling protein language models with vast, diverse data reveals emergent biological capabilities and enables programmable design.
• Researchers try to cut the genetic code from 20 to 19 amino acids — Researchers from Columbia and Harvard successfully engineered *E.
• Single-round evolution of RNA aptamers with GRAPE-LM — GRAPE-LM enables single-round RNA aptamer evolution using AI, outperforming multi-round human methods.
• Unlocking the Future of Gene Therapy — AI is revolutionizing gene therapy by enabling precise cellular targeting and accelerating discovery-to-clinic pathways.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.