FiberPro 1.0 is reported as a multiagent framework driven by a large language model that couples high-throughput fiber spinning with real-time experimental feedback. The system is intended to provide autonomous design, execution, and optimization for the production of functional protein micro- and nanofibers. The authors position FiberPro 1.0 as a solution to throughput and stability limitations that have historically constrained protein fiber manufacturing, and they describe a demonstration in which the platform is used both to find spinnable protein formulations and to deposit conformal, functional coatings on complex substrates.
Functional protein fibers combine high specific surface area with bioactive architectures, which make them attractive for a range of applications. However, the source describes two main obstacles: low production throughput and severe processing instability when spinning protein formulations. Focused rotary jet spinning (FRJS) is highlighted as a promising high-throughput technique that can enable direct, conformal deposition onto complex and irregular substrates. At the same time, FRJS operates within narrow processing windows for protein materials, and navigating those windows experimentally without closed-loop guidance is described as failure-prone.
FiberPro 1.0 unites multiple components into an autonomous workflow. Central to the system is a large language model–driven multiagent architecture that integrates experimental measurement and feedback into iterative design cycles. In practice, the platform proposes formulations or processing parameters, executes high-throughput spinning via FRJS, receives real-time experimental feedback, and then refines subsequent proposals. The source frames this closed-loop approach as a way to traverse the narrow and failure-prone parameter spaces associated with protein spinning using FRJS.
The authors report that FiberPro 1.0 was applied to three distinct protein systems. For each system, the platform carried out iterative optimization with the closed-loop experimental feedback. Across these three systems, FiberPro converged on spinnable formulations with an average of two iterations. The source presents this outcome as evidence of rapid optimization capability and as a demonstration of the system’s efficiency in identifying workable processing conditions within tight FRJS parameter windows.
To illustrate a translational application and the platform’s conformal deposition capability, FiberPro designed a zein-based active packaging system. The framework was used to deposit this zein formulation directly onto diverse food matrices via FRJS, producing a conformal coating that adhered to irregular substrate geometries. This demonstration emphasizes the system’s potential for real-world manufacturing and application scenarios where substrates are non-planar and conventional coating methods may struggle to achieve uniform coverage.
In the zein-based packaging demonstration, the conformal coating reportedly combined antibacterial activity with real-time freshness monitoring. The authors state that the resulting coating markedly suppressed Escherichia coli and extended shelf life of the coated food matrices. These reported functional outcomes are presented as evidence that autonomous AI-guided design, when integrated with high-throughput FRJS processing, can yield functional protein micro/nanofiber products with practical utility.
The study frames FiberPro 1.0 as connecting autonomous AI reasoning with high-throughput processing, thereby establishing a verifiable route for scalable manufacturing and conformal coating of functional protein micro/nanofibers. By reducing iteration counts to an average of two for finding spinnable formulations across three protein systems, the platform is presented as a tool to accelerate formulation discovery and to stabilize processing outcomes within the narrow windows required by FRJS. The zein-based packaging use case is offered as a concrete example of how this approach might translate to applied product development, particularly for conformal coatings on complex substrates.
The source is a preprint posted on bioRxiv with the listed authors and a DOI. The authors declared no competing interests. The preprint is made available under a CC-BY 4.0 International license. The source includes author affiliations and contact information for correspondence; additional experimental details, datasets, and supplementary material are referenced in the original preprint but are not reproduced here.