The Application of Reference Dose Prediction Model to Human Health Water Quality Criteria and Risk Assessment

Abstract

Oral reference dose (RfD) is a key parameter for deriving the human health ambient water quality criteria (AWQC) for non-carcinogenic substances. In this study, a non-experimental approach was used to calculate the RfD values, which explore the potential correlation between toxicity and physicochemical characteristics and the chemical structure of pesticides. The molecular descriptors of contaminants were calculated using T.E.S.T software from EPA, and a prediction model was developed using a stepwise multiple linear regression (MLR) approaches. Approximately 95% and 85% of the data points differ by less than 10-fold and 5-fold between predicted values and true values, respectively, which improves the efficiency of RfD calculation. The model prediction values have certain reference values in the absence of experimental data, which is beneficial to the advancement of contaminant health risk assessment. In addition, using the prediction model constructed in this manuscript, the RfD values of two pesticide substances in the list of priority pollutants are calculated to derive human health water quality criteria. Furthermore, an initial assessment of the health risk was performed by the quotient value method based on the human health water quality criteria calculated by the prediction model.

Systematic approaches to machine learning models for predicting pesticide toxicity

“…Exposure to pesticides can lead to various health issues in the reproductive system, fetal development, and an increased risk of cancer and asthma [ 7 , 11 , 25 , 30 ]. Moreover, the accumulation and persistence of pesticides in the environment makes them a chronic threat to human health [ [30] , [31] , [32] ].…”