Integrating Computational Optimization with Antimicrobial Susceptibility Testing: A Particle Swarm Optimization Framework for Enhancing Fluoride Toothpaste Formulations

This study develops a methodological framework that combines conventional antimicrobial susceptibility testing with Particle Swarm Optimisation (PSO) to enhance toothpaste formulations, employing Escherichia coli isolated from the oral cavity as a model organism. We used the agar well diffusion method to see if two fluoride toothpastes (Oral B and My-my) could kill oral E. coli isolates at 6.25%, 12.5%, 25%, 50%, and 100% concentrations. A surrogate Random Forest model was created using these experimental data to link formulation parameters to antimicrobial activity. Then, PSO was used to find the best formulation traits. Multi-objective optimisation that looks at the trade-offs between antimicrobial effectiveness and cytotoxicity was shown as a conceptual framework. Both toothpastes showed antimicrobial activity that depended on the concentration, with Oral B being more effective (23.0 mm at 100% concentration) than My-my (20.0 mm). The PSO framework, utilised as a methodological illustration while explicitly recognising data constraints, determined hypothetical formulation parameters (sodium fluoride 1100 ppm, hydrated silica abrasive, 2.5% SLS) with an anticipated zone of inhibition of 26.3 mm. These predictions are mathematically optimal for a surrogate model that was trained on very little data (n=10 formulation points). They need a lot of experimental testing before any claims about the formulation can be made. This work is presented as a proof-of-concept methodological framework, not as validated formulation guidance.