Enhancing the predictive capability of a mathematical model for pseudomonas aeruginosa through artificial neural networks

This paper talks about a type of bacterium called Pseudomonas Aeruginosa (PA) that can infect vulnerable people and can affect hospitals despite years of research. Its distinctive trait that increases the bacteria's infectiousness is because of Quorum Sensing (QS). It enables the bacteria to avoid detection by the immune systems of the hosts and to only start harming people when their colonies reach uncontrollably high populations. It is also well-recognised that a wide range of internal and external variables can influence QS. Existing studies show that a mathematical model has been proven to be successful in describing the QS trait, which could be modified to include coefficients that would capture the variability of this trait. However, we also believed that using Artificial Intelligence (AI)-based model to predict this trait, may provide a better understanding of it. Therefore, we apply Artificial Neural Networks (ANN) to estimate the Ordinary Differential Equation (ODE) coefficients of the mathematical model in a multi-output regression problem which results in a more accurate prediction about how the bacteria would behave under various variables. Results demonstrate that the mathematical model for PA was improved with this new information. Overall, the result shows that an ANN achieved a mean concentration error of 0.0637 that could predict 4 out of 11 coefficients with a range of \(\pm\) $\pm$ 40% from their original values mentioned in the literature. Through the potential use of ANN in diagnosis and treatment planning, the enhanced model's predictive capabilities give it the potential to save lives in clinical applications.