2.1 Why Synthesize Rare Ginsenoside Rh1?

As the main active components of ginseng, rare ginsenosides are a class of tetracyclic triterpenoid saponins with significant pharmacological activities. Based on differences in aglycone structures, dammarane-type ginsenosides can be divided into two major categories: protopanaxadiol (PPD) and protopanaxatriol (PPT) [5].

Among them, rare ginsenoside Rh1 has unique and significant advantages in the cosmetics field and is a core active ingredient in high-end anti-aging products. Studies have shown that at a concentration range of 10 ng/mL to 1 μg/mL, it can stimulate type I collagen synthesis by 50% to 160%, effectively promoting skin repair. Meanwhile, it can inhibit tyrosinase activity (IC₅₀ = 10.86 μg/mL) [1], reducing melanin deposition [3]. Its antioxidant and anti-inflammatory capabilities are far superior to those of traditional ginsenoside Rb1, which can meet consumers' dual demands for "potent anti-aging effects" and "safety and naturalness," demonstrating excellent performance.

2.2 Why Choose Saccharomyces cerevisiae as the Chassis?

Saccharomyces cerevisiae has become a highly promising platform cell factory and exhibits significant advantages in the field of biological manufacturing. As a chassis, it is first and foremost the first eukaryote with a fully sequenced genome, featuring clear and accessible genetic information that facilitates precise genetic manipulation. Secondly, it has obtained FDA safety certification and is approved for use in the production of biological products and food, ensuring reliable safety. Importantly, it possesses an endogenous mevalonate (MVA) pathway for terpenoid synthesis, providing a foundation for the synthesis of specific metabolites. Additionally, the P450 membrane proteins involved in saponin synthesis are more likely to achieve compatible expression in eukaryotes like Saccharomyces cerevisiae. Its complex organelle structure endows it with unique advantages in the post-translational modification of enzymes, contributing to the production of more functionally active products [6].

2.3 Why Choose Red Algae as the Carbon Source?

Red algae, as a carbon source for ginsenoside Rh1 synthesis, offers significant advantages in multiple aspects. Red algae have a short growth cycle, high photosynthetic efficiency, and do not require fertilization. They have a low lignin content (< 10%) and a high polysaccharide content of up to 61-67 wt%, mainly composed of agar polysaccharides [4], with an enzymatic hydrolysis conversion rate of up to 80.3%. In coastal areas, red algae residues accumulate heavily, resulting in almost zero pretreatment costs. Compared with the traditional glucose fermentation method, the total cost can be reduced by 47.24%. Furthermore, the monosaccharides produced by the hydrolysis of agar polysaccharides are galactose and 3,6-anhydro-L-galactose. Among them, the ethanol yield during galactose fermentation is only 0.27 g/g, much lower than the 0.39 g/g yield from glucose. This can maintain mitochondrial integrity to support the synthesis of high-energy-consuming terpenoids, avoid controversies related to food security, align with the concept of "carbon neutrality," and achieve a closed loop of sustainable production [2].

2.4 Design of Our Metabolic Pathway

With "sustainable utilization of marine resources" as the starting point, this project aims to develop high-end anti-aging cosmetics with Rh1 as the core component, focusing on the dual differentiated advantages of "high-purity de novo synthesis" and "zero arable land dependence." The project deeply integrates the "top-down" systematic design thinking and "bottom-up" component optimization logic of synthetic biology to precisely regulate the biosynthetic process and achieve efficient and green production of Rh1.

3.1 Why was FBA chosen as the feasibility analysis tool for this project?

This project aims to utilize red algae extract as a non‑traditional carbon source to overcome potential limitations associated with glucose metabolism. To evaluate the cellular uptake and conversion efficiency of unconventional carbon sources and their impact on the biosynthetic pathway of ginsenosides, quantitative analysis of metabolic flux is required. The core scientific question this project seeks to address is whether red algal hydrolysate, as a mixed carbon source, offers significant advantages in theoretical yield and sustainability compared to traditional single carbon sources like glucose.

Accurate quantification of metabolic flux typically relies on ¹³C‑labeled metabolic flux analysis (¹³C‑MFA). This method tracks the distribution of ¹³C‑labeled substrates within the metabolic network, detects isotopic labeling patterns using mass spectrometry or nuclear magnetic resonance techniques, and employs computational models to fit and obtain intracellular flux distributions. Although ¹³C‑MFA yields precise results, it is costly and involves a lengthy experimental process that includes cell culture, sample preparation, data acquisition, and modeling analysis.

To rapidly validate the feasibility of this project, we adopted flux balance analysis (FBA), a more efficient method that offers reasonable accuracy. FBA is a constraint‑based mathematical modeling approach used to predict flux distributions of metabolites in metabolic networks. Based on mass conservation and steady‑state assumptions, it constructs a system of constraints through stoichiometric relationships and optimizes flux distributions using an objective function (e.g., biomass maximization). This study employs the consensus genome‑scale metabolic model of Saccharomyces cerevisiae, GEM‑Yeast9, which includes 3,928 reactions, 2,666 metabolites, and 1,133 genes, and integrates 163 condition‑specific single‑cell models (csGEMs) and 1,229 strain‑specific models (ssGEMs). This model can incorporate constraints based on single‑cell transcriptomic and proteomic data, enabling accurate prediction of metabolic reprogramming under osmotic stress and nitrogen source limitation.

At the tool level, we use the COBRApy framework, which effectively represents complex biological metabolism and gene regulation processes. It features a robust open‑source ecosystem, high usability, and strong integration capabilities. By incorporating metabolic models validated through large‑scale experimental data, FBA can provide relatively reliable predictions of metabolic fluxes.

3.2 2.Why was FBA chosen as the software core and why was an Agent introduced?

Currently, there is a lack of dedicated FBA computational software with visualization and graphical interactive capabilities. Most tools still require users to perform operations through programming, and visualization is typically limited to final result presentation, while modeling and computational processes still involve coding. In traditional research settings, conducting metabolic network analysis demands that researchers possess both biological expertise and programming skills (e.g., using MATLAB with the COBRA Toolbox or Python with COBRApy), including knowledge of syntax, environment configuration, and debugging, which presents a high barrier to entry.

Our software achieves full‑process computational visualization and graphical interactive operation, significantly reducing the barrier to use. Users can complete all modeling steps via a web interface without writing command‑line code. Additionally, by leveraging the Escher tool to generate interactive metabolic flux maps, flux distributions become immediately clear, greatly enhancing the intuitiveness of result interpretation and research efficiency.

To further lower the barrier for users, we developed a specialized Agent based on DeepSeek‑R1:7B. Users only need to provide natural language instructions, and the Agent can invoke the corresponding tools to perform software operations, thereby eliminating the reliance on coding. Moreover, the system integrates a dedicated knowledge base for FBA. This knowledge base incorporates the practical experience of five synthetic biology researchers, systematically collating common issues, frequent operations, and typical application scenarios, ensuring that the Agent can more accurately understand user intent and provide effective responses.

The approach exemplifies a new human–computer interaction paradigm that can be extended to various research software in synthetic biology, systems biology, and biotechnology. By abstracting technical operations into conversational exchanges, such agent-assisted systems not only enhance accessibility but also accelerate iterative design-validation cycles. They provide scalable and adaptable support for a range of tasks—from metabolic network debugging to genetic parts assembly—enabling more researchers to translate innovative ideas into computable workflows with ease. This represents a step toward more democratized and widely accessible computational research infrastructures.