Conversational wet-dry closed loop, an protein experiment automation platform is defining the future

When biotechnology meets AI, the traditional research model of 'trial and error' is being redefined by an 'integrated' transformation.

From protein design computation, to automated wet lab validation, and then to data feedback driving the next round of AI iteration — today, more and more biomanufacturing companies are embedding these seemingly separate three links into a seamlessly connected 'integrated' platform.

This is the essence of an integrated intelligent biomanufacturing platform: AI design, third-party experimental services, expert services.

01 What is an protein experiment automation platform

Integrated Bio-Smart Manufacturing Platform

The core of an integrated bio-intelligence manufacturing platform is not a dazzling stack of hardware, but the integration of three types of scarce capabilities into a service that ordinary researchers can access:

AI design capabilities: Quickly generate high-performance sequences or pathways through protein design models, knowledge bases, and automation tools.

Third-party experimental services: Intelligently schedule external wet lab resources (such as plasmid synthesis, protein expression, and functional testing) without needing to build an in-house lab.

Expert services: Private services from domain experts, assisting with complex task decomposition and critical decision-making.

Based on the integration of these three elements, the platform delivers three key values:

Lowering the barrier — Individual developers or small teams can initiate high-level protein research without having a dedicated AI team, automation equipment, or lab personnel.

Shortening the cycle — The time from "generating an idea" to "obtaining verified results" is reduced from months or even years to just a few days.

Resource integration — AI design, third-party experiments, and expert services are scheduled and completed within a single conversation interface, so users do not need to coordinate with multiple vendors themselves.

02 How does the integrated platform work?

On April 24, 2026, in Shanghai, Tianhu Technology, an AI protein R&D company established only 5 years ago, officially released MatwingsVenus™ ™, a conversational protein R&D agent. This platform truly pushes the concept of integration into actionable product forms, extending the AI intelligence of the digital world to real-world laboratories.

The platform's business closed loop can be simplified into a clear four-step process: user dialogue → AI design → intelligent scheduling third-party experiment verification → data reflow

User dialog ordering: Users describe their R&D goals in natural language (e.g., "Design a sweet protein with better heat resistance").

AI design: The platform's built-in agent automatically disassembles tasks, calls protein design tools, databases, and prediction models to generate candidate sequences.

Intelligent scheduling of third-party experimental verification: AI automatically distributes candidate sequences to third-party experimental service providers (such as plasmid construction, protein expression purification, and activity detection) and tracks the execution status.

Data Reflow: Experimental results are structured backfeed to the platform for iterating on the next round of AI design, forming a self-optimizing wet and dry loop.

This model encapsulates the traditional "design-outsource-wait-redesign" fragmented process into an end-to-end task that can be completed in a single conversation.

The essence of this architecture is to transform R&D capabilities that were previously only available to large scientific research institutions and leading enterprises into a "shared laboratory" that can be easily accessed by individual developers.

Practical Validation: Designing Binders from Scratch and Sweet Protein Modification

The platform strategy of MatwingsVenus™ (Xiaowu™) has been practically validated in two real projects:

· De novo design of immune regulatory receptor binders: This was an extremely challenging selection—there were no similar drug molecule references, the target is dominated by polar surfaces, and the natural ligand already has nM-level high affinity. Based on target structure and functional requirements, MatwingsVenus™ (Xiaowu™) automatically completes the entire computational process including scaffold screening, interface design, sequence optimization, and druggability prediction through its Agents, quickly generating high-quality binder design sequences. The binders produced by the automated experimental platform resulted in dozens of molecules showing clear in vitro cellular blocking activity, combining functional inhibition with high affinity potential.

· Complex multi-round modification of the sweet protein Monellin: Natural Monellin is highly sweet but unstable. MatwingsVenus™ (Xiaowu™) utilizes a multi-round iterative strategy of "Agent design—automated wet experiments—AI feedback—Agent redesign," narrowing the search space with each round. Eventually, several samples had sweetness increased by over ten times compared to the wild type while maintaining high thermal stability.

These two cases demonstrate that the integrated platform enables the "idea → validation → iteration" process to be completed in a few days to weeks, with all data automatically fed back to continuously optimize the AI model.

03 Platformization is reshaping the 'ecosystem concept' of biomanufacturing

The significance of an integrated intelligent biomanufacturing platform lies not only in improving the efficiency of individual enterprises but also in how it is reshaping the entire industry's capacity distribution. Block Field

Tech Enzyme Library Biotechnology Laboratory

The capital side also confirms the popularity of this sector. After Matwings Technology completed over 100 million yuan in Series A financing in November 2024, it again raised more than 200 million yuan in Series A financing in March 2025, led jointly by CNPC Kunlun Capital, existing shareholder Yonghua Investment, and Shanghai Future Industry Fund. Capital is betting not just on a company's algorithm, but on the systematic R&D infrastructure represented by the entire 'integrated platform' direction.

04 New Infrastructure, Timely Opportunity

Enzyme library

Market Scale: According to data from the Ministry of Industry and Information Technology, China's production of bio-fermentation products accounts for over 70% of the global total, with an industry scale reaching 1.1 trillion yuan. CCID Consulting predicts that by 2026, the scale of the synthetic biology manufacturing industry will surpass 100 billion yuan.

Policy Support: In February 2026, eight departments including the Ministry of Industry and Information Technology issued the "Implementation Opinions on the Special Action for 'Artificial Intelligence Manufacturing'," listing biomanufacturing as a key area. In April the same year, the Ministry of Industry and Information Technology and the National Data Bureau launched the "Modular Resonance" initiative, explicitly promoting the collaborative interaction of AI models and industry data.

Capital Focus: After Matwings Technology completed over 100 million yuan in Series A financing in 2024, it secured more than 200 million yuan in Series A financing again during 2025-2026, led jointly by CNPC Kunlun Capital, Yonghua Investment, and Shanghai Future Industry Fund.

These signals together point to one judgment: the next phase of competition in biomanufacturing will no longer be a contest of individual algorithms or single devices, but whether it is possible to build an integrated platform that brings together AI design, third-party experiments, and expert services—allowing every idea to be verified with low barriers and high efficiency, and turning every verification into fuel for driving the next generation of AI.

An integrated bio-intelligent manufacturing platform may very well be the most noteworthy new infrastructure in this wave of construction.