Machine Learning-Based Prediction of Bacterial Growth in Laboratory Cultures (Predicting Colony Growth Rate Based on Temperature, Ph, and Nutrient Concentration)

Abstract

Accurate prediction of bacterial growth rates is essential for optimizing laboratory experiments, industrial bioprocesses, and food safety monitoring. Traditional mechanistic models, while interpretable, often struggle to capture the non-linear and interactive effects of environmental variables such as temperature, pH, and nutrient concentration. This study evaluates the performance of two machine learning models Random Forest (RF) and Support Vector Regression (SVR) in predicting bacterial growth rates using a curated dataset of 200 experimental observations.
Data preprocessing included feature standardization, and model evaluation employed a hold-out validation approach with metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), and R² score. Results demonstrate that the RF model substantially outperformed SVR, achieving an R² of 0.853, MAE of 0.0542, and MSE of 0.0205, compared to SVR’s R² of 0.587, MAE of 0.1354, and MSE of 0.0574.
Feature importance analysis revealed that temperature was the dominant determinant of growth (63%), followed by pH (31.1%) and nutrient concentration (5.9%). Visualization via scatter plots, grouped bar charts, and heatmaps confirmed the superior predictive accuracy of RF and highlighted the nonlinear growth responses characteristic of bacterial cultures. The findings indicate that RF-based models provide a robust, data-driven framework for predicting microbial growth, reducing experimental workload, and guiding optimal culture conditions in laboratory and biotechnological applications. Limitations include strainspecificity and the black-box nature of the model, which does not explicitly account for underlying metabolic mechanisms.