Machine learning prediction of allelochemical inhibition ratio (IR) on Microcystis aeruginosa growth

Abstract

This study consolidates a dataset of 83 allelochemicals derived from plant parts, microbes, marine organisms and agricultural byproducts, underscoring their ecological relevance. While allelopathic inhibition is a promising process for suppressing HABs, current research exhibits inconsistencies in elucidating its underlying mechanism. Furthermore, the field remains constrained by a lack of systematic comparisons among machine learning (ML) models. To bridge these gaps, we integrated multi-source experimental data and employed diverse ML algorithms to predict allelochemical inhibition ratio (%) against Microcystis aeruginosa. This methodology pinpoints critical variables for optimizing treatment efficacy and addresses key limitations of conventional experimental approaches. Through quantitative analysis, novel insights into feature importance and inhibition dynamics are established, significantly advancing predictive accuracy in HAB management. Statistical analysis of the dataset revealed an average inhibition ratio of 54 %, with values ranging from 10 % to 100 %. Among evaluated ML models, Random Forest, Bagging, and Extreme Gradient Boosting Regressors showed superior performance in predicting inhibition ratio (%). From a dataset of 83 allelochemicals classified into 15 categories, exposure concentration, exposure time, algal biomass, linoleic acid microspheres, flavonoids, and plant extracts were identified as most influential factors affecting inhibition efficiency. Notably, Microcystis aeruginosa inhibition was highly sensitive to shorter exposure durations (10–12 days), lower algal biomass (optimal at 1 × 10⁷ cells/mL), and concentrations exceeding 0.05 g/L. The exponential growth phase emerged as a critical window for bloom controlling. Overall, this framework offers a data-driven foundation for policy-makers and researchers to design targeted biocontrol strategies, mitigating the ecological and economic threats posed by HABs.