Limitations of the refolding pipeline for de novo protein design 1. A new study critically evaluates the widely-used refolding pipeline in de novo protein design, revealing significant limitations that could impact the reliability of sequence-structure compatibility assessments. The research highlights how evolutionary information can obscure the predictive power of refolding metrics like pLDDT and scRMSD, especially for sequences homologous to natural proteins. 2. The study demonstrates that including evolutionary information in structure prediction models, such as using MSAs with AlphaFold2, reduces their ability to distinguish between well-designed and poorly designed sequences. This finding challenges the common practice of using MSAs in protein design campaigns and suggests that single-sequence mode might be more reliable for certain design tasks. 3. The authors systematically analyzed the impact of different prediction settings on designability metrics. They found that flexibility and disordered regions in protein structures can significantly affect the accuracy of scRMSD, leading to inflated values and unreliable designability assessments. A simple realignment correction was proposed to mitigate this issue. 4. The research also explores the influence of sequence length and secondary structure content on designability, showing that longer proteins and those with mixed secondary structures are more challenging to assess using current refolding pipelines. This highlights the need for improved metrics and methods to better evaluate de novo protein designs. 5. The study concludes that while refolding metrics have been instrumental in protein design, their limitations must be carefully considered. The findings call for a reevaluation of current best practices and the development of new approaches to enhance the reliability of computational protein design tools. 📜Paper: biorxiv.org/content/10.64898… #ProteinDesign #ComputationalBiology #DeNovoProteins #RefoldingPipeline #Limitations #AlphaFold2