In August 2026, a paper published by Duke University and Harvard Medical School teams on the preprint platform bioRxiv is a brilliant testament to this new paradigm. The PRAXIS system they established gave AI the ability to conduct autonomous experiments and interact for the first time.

AI autonomous experiments close the loop, opening a new paradigm in protein engineering

At the core of this system lies a tightly coupled closed loop: the generative protein-language model ProteinNPT acts as the AI brain, while the fully automated robotic lab acts as the precise hands. AI designs enzyme variant sequences, robots automatically perform gene assembly, cell-free protein expression, and biochemical activity detection, and the resulting experimental measurements are then transmitted in real time to AI to update internal models and guide the next round of design. The entire cycle from design to obtaining functional data takes only about 10 hours, meaning AI can complete multiple iterations of "design-experiment-learn" in a short time, continuously improving like a tireless scientist.

At the system architecture and experimental level, the research team targeted glycoside hydrolase (GH1) from the first family and deployed three independent AI agents to explore substrate specificity for glucose, xylose, and mannose. These three agents not only make decisions independently but also share a single central experience database. Surprisingly, during the optimization of preset substrate activity, the system accumulated a large amount of real mutant test data, allowing the model to spontaneously learn patterns of "protein expression levels" and deep conformational changes at active sites.

To fully bridge the last mile from "AI computing power design" to "real data collection," GentleGen has specially developed the [High-Throughput Upper Definition Expression Service] for high-throughput screening! We do not advocate unnecessary "overpurification" during the initial screening stage, but rather provide you with high-throughput, highly parallel, high-quality supernatant samples through automated and standardized systems.

Full-process automation makes AI high-throughput screening a routine in the lab

Relying on its independently developed GentSyn-Exp3.0G-Lab fully automated production line, GentleGen has built a one-stop, high-throughput technology platform covering gene design, synthesis, validation, protein expression, and antibody affinity characterization. Traditional R&D models rely heavily on manual validation, but GentleGen has redefined the efficiency and cost of wet experiments through automation.

• Cost-performance revolution: High-throughput expression screening greatly reduces the detection cost of individual samples—IgG monoclonal antibody at 48.9 USD/clones, VHH nanobody at 28.9 USD/clones. Previously, the budget only allowed for synthesizing one or two antibodies, but now GentleGen can complete high-throughput screening of dozens of antibodies.

• Ready to use: Each antibody comes with 800 μl of antibody supernatant for direct functional validation—eliminating purification steps and skipping intermediate steps, delivering one-stop downstream testing from genes to antibodies.

What AI fears most is the "invalid feedback" caused by failed experiments. GentleGen leverages automation and AI algorithm optimization to achieve a first-time synthesis success rate as high as 99%, with sequence accuracy of 99.9%. This ensures that every feedback signal received by AI agents is genuine and reliable, greatly improving the efficiency of autonomous exploration.

The Duke University team demonstrated the feasibility of "AI through interactive learning" with PRAXIS, while GentleGen is transforming this cutting-edge paradigm into a new quality productive force accessible and reusable in China's bioeconomy through a robust automation platform.

For detailed information about related services, please send an email to marketing@gentlegen.com.