Overview

Within the POLARTO(Practical Optimization with Linked Affinity, Reaction and Transmission Oversight) modeling framework, our efforts are structured around four central research questions derived from our experimental and hardware design, as well as practical implementation scenarios:

  1. Is each reaction step in the proposed detection pathway thermodynamically and kinetically feasible?
  2. How to predict the binding strength of peptides based on sequences and propose optimization suggestions for the coiled-coils domain (CC domain) design in the project?
  3. How can our detection strategy effectively prevent the spread of Pseudomonas syringae pv. tomato (i.e., P. s. pv. tomato, or Pst), and what framework maximizes its economic benefit?
  4. How to ensure that our hardware is feasible for reliably detecting target signals?

To address these questions, we developed a series of interdisciplinary models integrating molecular dynamics, artificial intelligence, chemistry, physics, and epidemiology.

  • For the first question, we employed GROMACS to rigorously evaluate the thermodynamic and kinetic feasibility of each reaction step.
  • For the second, we constructed Seq2Affinity, a computational prediction and design model, to provide suggestions for optimizing the CC domains in our system and build a platform for broader estimation and design tasks.
  • For the third question, we developed a CS model (Cellular automaton based on the SEIR framework) to assess the epidemiological impact of early detection and applied Q-Learning to identify economically optimal detection strategies.
  • For the fourth, we utilized COMSOL Multiphysics to establish the relationship between glucose concentration and output current, thereby validating and guiding the hardware design.
  • In addition, we also adopted ordinary differential equations to simulate the whole system and estimate the time to reach equilibrium.

For more details, see the description below.

Molecular Dynamics

We simulated the behavior of peptides and proteins under near-physiological conditions (0.15 M ionic strength, 1 bar pressure) using GROMACS, to examine the binding affinity, energetic stability, and structural conformation of protein complexes involved in our signaling pathways.

Show More

Seq2Affinity: A Prediction and Optimization System for Peptide Binding

Seq2Affinity is a computational model we developed for rapid prediction and optimization of peptide binding affinity, mainly for CC domains. We built a specialized database and trained a machine learning model combining ESM-2 embeddings with LASSO regression, effectively handling high-dimensional data with limited samples. The framework also provides two mutation design strategies: exhaustive search for single-site mutations and genetic algorithm for multi-site optimization, and adopts the predictive model as the fitness criterion.

Show More

Epidemiological & Economic Benefit Model

This component creatively employs a coupled CA-SEIR model, incorporating environmental water and soil systems to develop an epidemiology-based framework that spans from plants to the environment, thereby visualizing transmission dynamics and infection outcomes. Building upon this foundation and focusing on the core objective of Pst detection, we designed a deep learning model centered on Q-learning.

Show More

Glucose-Potential Simulation for Hardware

This component simulates the enzymatic reaction of glucose and the corresponding potential changes within the chamber using COMSOL, establishing a concentration gradient model of glucose near the electrode. The resulting glucose-potential correlation curve theoretically validates the feasibility of the hardware group's detection tool, while also providing relevant threshold data.

Show More

Dynamic Reaction Analyzer -ODEs

To visualize the kinetics of our detection system and estimate the equilibrium time of the system, we established a set of ordinary differential equations (ODEs). The simulation results demonstrated an ~800-second equilibration for the GFP-output pathway, whereas the hardware trehalase pathway achieved a rapid trehalose concentration peak at 60 seconds, with full equilibrium in about 15 minutes.

Show More

© 2025 - Content on this site is licensed under a Creative Commons Attribution 4.0 International license.

The repository used to create this website is available at gitlab.igem.org/2025/ucas-china.