Abstract:Can one design and automate a computational and experimental platform such that each platform iteratively guides and drives the other to achieve a pre-determined goal? Rapp and colleagues (2024) describe just this possibility in a paper that details a prototype of a self-driven laboratory that can navigate autonomously to yield an engineered enzyme with a desired attribute. This laboratory, rather, the automated protocol, is referred to by an acronym - SAMPLE. This refers to Self-driving Autonomous Machines for Protein Landscape Exploration. The paper describes a prototype involving the engineering of a glycoside hydrolase for enhanced thermostability. The 'brain', the computational component behind this automated system, was designed to learn protein sequence- function relationships from a curated dataset. These designer proteins were then evaluated by a fully automated robotic system that could synthesize and experimentally characterize the designed protein and provide feedback to the agent, i.e., the computational component, to fine-tune its understanding of the system. The SAMPLE agents were thus designed to continually refine their understanding of the protein landscape by actively acquiring information in the search process. As this intelligent agent learns protein sequence-function relationships from a curated, diverse dataset, this feedback is crucial to refine landscape exploration and the design of new proteins based on the updated hypothesis. In this prototype, four SAMPLE agents were tasked with this goal. The goal of each of these agents was to navigate the glycoside hydrolase landscape and identify enzymes with enhanced thermal tolerance. Differences in the search behavior of individual agents primarily arise from experimental measurement noise. However, despite differences in their search behavior, all four agents could converge on a thermostable glycoside hydrolase - a remarkable feat as it apparently did not need any human intervention.

Self-driving autonomous machines to engineer novel enzymes

Accelerated enzyme engineering by machine-learning guided cell-free expression

Ultrahigh-Throughput Enzyme Engineering and Discovery in In Vitro Compartments

Revolutionizing enzyme engineering through artificial intelligence and machine learning

The road to fully programmable protein catalysis

'Set it and forget it': automated lab uses AI and robotics to improve proteins

In Vitro Enzyme Self-Selection Using Molecular Programs

Enhanced Sequence-Activity Mapping and Evolution of Artificial Metalloenzymes by Active Learning

On synergy between ultrahigh throughput screening and machine learning in biocatalyst engineering

Accelerating Biocatalysis Discovery with Machine Learning: A Paradigm Shift in Enzyme Engineering, Discovery, and Design

Microdroplet screening rapidly profiles a biocatalyst to enable its AI-assisted engineering

Computation-guided engineering of distal mutations in an artificial enzyme

Harnessing generative AI to decode enzyme catalysis and evolution for enhanced engineering

Protein Engineering Design from Directed Evolution to De Novo Synthesis

Autonomous chemical science and engineering enabled by self-driving laboratories

Generative Enzyme Design Guided by Functionally Important Sites and Small-Molecule Substrates

Ensemble-based enzyme design can recapitulate the effects of laboratory directed evolution in silico

Protein engineering via Bayesian optimization-guided evolutionary algorithm and robotic experiments

Engineering of highly active and diverse nuclease enzymes by combining machine learning and ultra-high-throughput screening

Combining chemistry and protein engineering for new-to-nature biocatalysis