📝 Quick Start & Notes

1. Input protein structure

Upload a PDB file or enter a PDB ID / UniProt ID.
Structures are automatically retrieved from RCSB PDB or AlphaFold DB.

2. Structure requirements

Only single-chain proteins are currently supported.
Multi-chain complexes and protein assemblies are not supported in the public server version.

3. Finetuning option

Enable finetuning to perform lightweight meta-transfer learning for protein-specific adaptation.
Typical finetuning takes ~10 min depending on server workload.
During finetuning, Foldseek homolog structures longer than 600 residues are automatically excluded from the training dataset.

⚠️ Warning:
Fine-tuning of proteins larger than 600 residues is not recommended on the public server due to computational resource limitations.

For longer proteins, please use recommended alternatives: Colab notebook · Local software from GitHub

4. Output results

high-fitness mutations
interactive protein structure visualization
mutation fitness landscape analysis

💡 What can unZipro do?

Enzyme engineering
Genome editing tools (SpCas9, Cas12, base/prime editors)
Plant protein engineering (virus resistance, transcription factors)
Protein therapeutics
General protein design in biotechnology & agriculture

🧪 Experimentally validated examples

Category	Description
Genome editors	Adenine base editor (TadA8e)
Genome editors	Three nucleases (SpCas9, CasΦ2/Cas12j2, T5E)
Polymerase	MMLV reverse transcriptase under multiple conformations
Fluorescent enzyme	Luciferase for improved fluorescence intensity
Plant proteins	DNA-binding domains of plant transcription factors
Plant proteins	Plant virus-resistance (R) proteins

📖 References

Please cite unZipro if used in research.

Qin, Z., Zhao, S., Deng, Z., Si, X., et al. (2026). Simplifying in silico protein evolution with minimal screening by unZipro. Molecular Cell. https://doi.org/10.1016/j.molcel.2026.08.028