How to Use NeuroFold2

NeuroFold2 is an enzyme-design workflow for generating variants that balance catalytic function with developability properties such as stability and solubility. Relative to sequence-only optimization, the Neurosnap implementation is explicitly conditioned on a reference enzyme structure and active substrates, making it better suited to active-site engineering and chemistry-aware enzyme redesign.

On Neurosnap, researchers provide a Reference Structure, one or more Active Substrates, optional Fixed Positions, and a target Optimal pH. This setup is useful when catalytic residues, cofactor contacts, or substrate-recognition motifs must be preserved while the rest of the enzyme is explored for improved performance.

The underlying NeuroFold work describes a multimodal enzyme-design strategy that combines generative modeling with physics-based and evolutionary signals to navigate sparse functional landscapes around a starting enzyme. NeuroFold2 applies that idea in a more structure-aware setting: the model sees the reference fold and substrate context, so proposed sequences are judged not only by generic sequence plausibility but also by whether they remain compatible with the catalytic scaffold.

On Neurosnap, Fixed Positions is the key control when certain residues cannot move, for example catalytic side chains, metal-coordinating motifs, or interface contacts. Because substrates are included directly, the workflow is better aligned with biocatalysis and active-site redesign than generic protein optimization tools that operate without chemical context.

Returned candidates are most useful when compared across several signals at once. Overall quality, sequence fitness, structural fitness, solubility, thermostability, optional kcat estimates, and structural clustering help separate conservative improvements from more radical but higher-risk variants.

• Catalysis-aware inputs: NeuroFold2 accepts a reference structure plus active substrates, so redesign stays anchored to the actual chemistry of the enzyme system.
• Residue-level control: Fixed Positions makes it practical to preserve catalytic motifs, binding residues, or structurally essential regions while redesigning the rest.
• Multi-objective screening: Activity-related, structural, and developability signals are reviewed together instead of optimizing only one property at a time.
• Campaign-ready interpretation: Ranked candidates and structural clustering help researchers choose whether to test conservative variants first or explore more diverse sequence families.