Presentation Information
[P04-456]Machine Learning–Assisted Enzyme Engineering of Phosphoketolase
○Paul P. Lin1, Emma J. Huang1, Dong-Ge Yu1, Chih-Hsuan Wang1, Jinn-Moon Yang1 (1. National Yang Ming Chiao Tung University (Taiwan))
Keywords:
Directed evolution,Protein engineering,Machine Learning,Phosphoketolase
Directed evolution is a robust strategy for optimizing protein properties, typically utilizing random mutagenesis via error-prone PCR (epPCR). However, because the cumulative effects of mutations are often non-additive (epistatic), identifying superior variants frequently necessitates exhaustive, trial-and-error screening. To streamline this process and navigate the complex fitness landscape of phosphoketolase (F/Xpk), we implemented a machine learning (ML)-guided engineering framework.
In this study, a library of F/Xpk variants was generated using epPCR to introduce single amino acid substitutions. To efficiently sample the functional space, variants were expressed in an acetyl-CoA auxotrophic strain, where cell growth served as a high-throughput in vivo proxy for F/Xpk activity. This selection step effectively filtered out non-functional variants, providing a refined pool of active enzymes for biochemical characterization. Following protein purification and enzymatic activity assays, we identified 23 single-site variants that exhibited superior activity relative to the wild type (WT). Notably, four of these variants demonstrated at least a 30% increase in enzymatic activity over the WT. This high-quality sequence-activity dataset was subsequently utilized to train Evo-AttnMPMN, providing a foundation for predicting synergistic mutations and further accelerating the development of high-performance biocatalysts.
In this study, a library of F/Xpk variants was generated using epPCR to introduce single amino acid substitutions. To efficiently sample the functional space, variants were expressed in an acetyl-CoA auxotrophic strain, where cell growth served as a high-throughput in vivo proxy for F/Xpk activity. This selection step effectively filtered out non-functional variants, providing a refined pool of active enzymes for biochemical characterization. Following protein purification and enzymatic activity assays, we identified 23 single-site variants that exhibited superior activity relative to the wild type (WT). Notably, four of these variants demonstrated at least a 30% increase in enzymatic activity over the WT. This high-quality sequence-activity dataset was subsequently utilized to train Evo-AttnMPMN, providing a foundation for predicting synergistic mutations and further accelerating the development of high-performance biocatalysts.
Comment
To browse or post comments, you must log in.Log in