Quantifying pesticide residues in food matrices using statistical methods

Traditional techniques for pesticide residues detection and quantification in food using mass spectrometry often require the analysis of standards for each compound, leading to time-consuming and laborious procedures, especially considering that these methods usually involve hundreds of pesticides to be targeted. This paper presents a novel approach to compound quantitation, where multiple target compounds can be accurately quantified using few predictor compounds, significantly reducing the experimental time and minimizing resource requirements. For this purpose, data on detector response which encompassed calibration slopes of a total of 96 pesticide standards on GC-MS/MS and 66 standards on LC-MS/MS, over a period of 4 years, were collected. Two fundamental statistical techniques, Pearson correlation and linear regression, were used to create a predictive model, which accuracy was further evaluated using R-square, adjusted R-square, and Root Mean Square Error (RMSE). Four predicted compounds for the LC dataset and seven predicted compounds for GC dataset were identified, and various predictor combinations were considered. A linear regression model was developed using predictor combinations to estimate the calibration slope of each of the target compound. This model was applied for the quantitation of several pesticide residues in food samples. The results validated the model’s high accuracy.