[2S00a] 01. ASBA - Part 5

3:05 PM - 3:20 PM JST(6:05 AM - 6:20 AM UTC)

[2ASBA-11-TA] Mapping the Protein Fitness Landscape via Evolution, AI, and Industrial Automation

○Shuyi Zhang1 (1. Tsinghua University, China)

Video Presentation

Keywords:

1.Protein design,2.AI,3.Evolution,4.Automation

Video Presentation

Protein engineering is at the forefront of biotechnological innovation, offering the promise of designing proteins with unprecedented functionalities for various applications. However, current methods for engineering novel protein functions are limited by the lack of understanding of the sequence-function relationship, the difficulty in designing complex properties using AI, and the labor-intensive nature of traditional directed evolution methodologies. Here, we demonstrated that the protein sequence-function space is highly compressible (with a compression ratio of 10^48) through the development of a new evolution method, EvoScan, and a corresponding AI method, EvoAI. We have further developed an industrial-grade automation platform capable of high-throughput, efficient, and reliable evolution with minimal human intervention (uninterrupted operation for ~1 month). Additionally, we created a genetic circuit-controlled and growth-coupled continuous directed evolution system that enables the intelligent evolution of proteins with diverse and complex functionalities. These two systems were integrated into an industrial self-driving laboratory, iAutoEvoLab, where we demonstrated that the platform could evolve proteins from non-functional precursors to fully functional entities. The evolved proteins were then directly applied to in vitro mRNA transcription and mammalian systems. Our integrated methods of using evolution, AI, and industrial automation demonstrate significant potential for protein engineering and mapping the landscape, advancing both protein engineering and synthetic biology.