AI surpasses humans again? This time, it's about protein modification

Artificial intelligence (AI) has once again surpassed humans, this time in modifying proteins.

Tasks that might take human scientists 6-12 months to complete can be accomplished by AI in just weeks, without the need for human intervention, feedback, or subjective judgment.

This is an AI-driven, fully automated robot that can engineer proteins. SAMPLE was proposed by a research team from the University of Wisconsin–Madison and is a proof-of-concept for protein design and construction without human intervention.

The related research paper, titled "Self-driving laboratories to autonomously navigate the protein fitness landscape", has been published in the inaugural issue of Nature Chemical Engineering, a subsidiary of Nature.

The research team said that SAMPLE automates and accelerates the scientific discovery process and has important implications for the fields of protein engineering and synthetic biology.

AI intelligent body, transforms enzyme with better heat resistance

Protein is the material basis of all life on Earth and is involved in every process of cellular activity.

Protein design, which is the ability to create new proteins with specific functions and properties, has been widely used in fields such as biotechnology, chemistry and medicine, such as:

Develop new drugs and therapies, especially in the fields of cancer, cardiovascular disease, genetic disease treatment, etc., and also contribute to vaccine development and personalized medicine;

Creating enzymes and other biocatalysts for use in biomanufacturing to efficiently convert feedstocks into useful products such as biofuels, pharmaceuticals, or food ingredients;

Improve the disease resistance, drought resistance and nutritional value of crops, thereby improving agricultural productivity and food security;

Degrading pollutants in the environment, such as heavy metals and organic pollutants;

Constructing new biomaterials, such as biocompatible materials for medical engineering, or high-performance materials with unique physical properties;

It helps scientists gain a deeper understanding of the relationship between protein structure and function, and promotes the development of biochemistry, molecular biology and other fields. Although protein engineering has great application potential, creating a new protein with improved or new functions is still a repetitive and laborious process, which sometimes takes human scientists many years.

In the process of studying biological systems, scientists deepen their understanding of the system by generating hypotheses, designing experiments to verify the hypotheses, conducting experiments in the laboratory and interpreting the data, and then repeatedly iterate this process to gradually reveal biological mechanisms and design new systems with better performance and behavior.

Today, AI is used to automate tasks in all walks of life, but due to the complexity of biological traits and experiments, it remains a challenge to develop a fully automated system that can engineer proteins without rest and learn from the data generated.

According to reports, SAMPLE is driven by an AI agent that can learn the relationship between protein sequence and function, design new proteins, and send these proteins to a fully automated robotic system that experimentally tests the designed proteins and provides feedback, thereby improving the AI ​​agent's understanding of the system .

Figure｜SAMPLE, a fully autonomous protein engineering system.

To test the system, the researchers used four SAMPLE agents to engineer glycoside hydrolases with improved heat resistance . Despite their different search behaviors, each SAMPLE agent ultimately found an enzyme that was more thermostable, at least 12°C more stable than the original starting sequence.

Figure ｜ 4 SAMPLE agents autonomously discovered enzymes with better thermal stability.

A universal protein engineering platform

The research team said that SAMPLE is a general protein engineering platform that can be widely used in the fields of bioengineering and synthetic biology.

Although they only demonstrated the system's performance in thermostability engineering, the same approach could be used to modify enzyme activity, specificity, and even create chemical reactions not seen in nature .

Like directed evolution, the system requires no prior knowledge of protein structure or mechanism, but instead uses an unbiased approach to study how sequence changes affect function.

However, the research team also pointed out that the biggest obstacle to establishing SAMPLEs for new protein functions is the required biochemical detection . The robotic system used in this work can only use a microplate reader, so colorimetric or fluorescence-based detection methods are required.

In principle, more advanced analytical instrumentation, such as liquid chromatography-mass spectrometry or nuclear magnetic resonance spectrometry, could be integrated into automated systems, thereby expanding the types of protein functions that can be designed.

In addition, delays in obtaining resources, robot failures, and system downtime may also affect the total time it takes to develop proteins using the system .

The research team has now implemented the complete experimental workflow on the Strateos Cloud Lab, creating a cost-effective, easy-to-use system that can be adopted by other synthetic biology researchers.

In the future, autonomous laboratories such as SAMPLE will revolutionize the fields of biomolecular engineering and synthetic biology, automating inefficient, time-consuming and laborious protein engineering activities and allowing researchers to focus more on important downstream applications.

As deep learning, robotic automation, and high-throughput instrumentation continue to advance, intelligent automated systems for scientific discovery will become increasingly powerful.

Paper link:

https://www.nature.com/articles/s44286-023-00002-4