Machine learning prediction of copper ion interference with mercury ion fluorescence signals in food heavy metal detection

Abstract

Objective: To construct an artificial intelligence prediction model to predict the selectivity of fluorescent probes for Hg2+ in a complex food testing environment in the presence of Cu2+ interference. Methods: Fluorescent probe technology combined with seven advanced classical machine learning models was used to predict and analyze the selectivity of the probe for Hg2+ in the presence of Cu2+ interference, and to compare the prediction effect of each model and select the optimal model. Results: Efficient models with accuracies of 0.786 and 0.810 in the cross-validation and test sets were successfully established based on Molecular 2D Descriptors （Mol2D） and extreme gradient boosting algorithms to accurately predict the probe selectivity of Hg2+ under Cu2+ interference. Conclusion: The model is improved for the design of Hg2+ fluorescent molecular probes by selective prediction, which makes the design of Hg2+ fluorescent probes more efficient and reliable.

Keywords

• mercury ion detection
• fluorescent molecular probes
• probe selectivity
• machine learning
• cheminformatics