In the NAU-CHINA 2025 project, we conducted our work from three aspects: molecular modeling,mathematical modeling and fashion modeling. In molecular modeling, we constructed a quantitative model to explore the structure of the FourU switch at different temperatures, providing a theoretical basis for temperature regulation of gene expression. We also predicted the tertiary structure of the protein and conducted molecular dynamics simulations on it, which helps us better understand the structure and properties of the protein used and proves the rationality of the experimental design. In mathematical modeling, we visualized the role of the cascade system and, in combination with the data from E.coli co-culture experiments, determined the strain interaction types through statistical tests. With the goal of maximizing yield, we designed an interactive co-culture model. Meanwhile, for the staining process of pigments in bacterial cellulose membranes, we also constructed a PDE model to track and simulate the diffusion and adsorption of pigments. In fashion modeling, we combined machine learning algorithms to construct an intelligent size matching model with multi-level scoring. By comparing and visualizing the water-saving effects, we proved that the topic was of great significance. In addition, we also use iterative algorithms to generate fractal diagrams, adding new fashion elements to textile pattern design.

Figure 1 | Model General Flowchart

We conducted an in-depth analysis of the FourU temperature-sensitive riboswitch and the BslA-dCBM fusion protein. For FourU RNA, we conducted qualitative simulations using RNA Structure to observe the exposure of temperature-dependent SD sequences. On this basis, a quantitative model for thermodynamic and dechaining temperature prediction was further constructed. The predicted dechaining temperature calculated through Python was close to the design target value. For fusion proteins, we used AlphaFold to predict the high-confidence structures of individual domains, and referred to literature to analyze the functional integrity and stability of fusion proteins with the help of the Kabsch algorithm and GROMACS. We also evaluated the optimal number of CBM proteins from multiple dimensions such as binding sites, hydrophobicity, and structural feasibility.

Learn more about Molecular Modeling

We first explored the role of the cascade system with ODE and performed a sensitivity analysis of the parameters. Subsequently, we obtained data from the wet experimental group, extracted growth parameters and classified the interactions using hypothesis testing and interaction indicators. For the competitive relationship, the Lotka-Volterra model is adopted to calculate the competitiveness. For neutral or reciprocal relationships, quantitative analysis is conducted through relative growth in competitiveness and cooperation efficiency. Subsequently, we established a prediction model and used grid search to optimize the initial inoculation ratio and temperature switching time to achieve the maximum target yield. In addition, we have developed a PDE model that combines FEM and random particle tracking technology to simulate the diffusion or adsorption process of pigments in BC membranes and match the experimental saturation time.

Learn more about Mathematical Modeling

By comparing different Logo recognition algorithms, we selected YOLOv8 for identification and recognition. Meanwhile, the MediaPipe library was adopted to carry out human bone detection, and based on the principles of anthropometry and the matching degree formula, a dynamic size intelligent matching model supporting multi-level scoring was constructed. To quantify the design advantages, we compared the water consumption in the production of cotton and bacterial cellulose, and identified the core competitiveness of our design over traditional industries. In addition, we have generated the Mandelbrot/Julia set fractal graph in Python using an iterative algorithm, and combined it with Photoshop and AI for optimization simulation. This can be applied in the fields of fashion and home textiles, enriching design materials with a digital, high-precision, and short-cycle creation mode.

Learn more about Fashion Modeling