Prediction of association constants of cesium chelates based on Uniform Design Optimized Support Vector Machine

Abstract

Mushrooms that have demonstrated a high affinity for radiocesium uptake were edible. Humans eating those mushrooms accumulated radiocesium and have been driven by health concerns in recent years. Considering cesium chelates exhibited high stability constant, environmental concentration of radiocesium was diminished by macrocyclic compounds, and the method has been seen as one of the best way to reduce radiocesium pollution. To understand the factors of cesium chelates' stability and predict stability constant value of a new macrocyclic compound, quantitative structure–property relationship studies based on molecular descriptors were presented for building models of cesium chelates. The stability constant, namely the association constant (logK), was correlated with six descriptors by Best Multi-Linear Regression (BMLR), Radial Basis Function Neural Network (RBFNN) and Uniform Design Optimized Support Vector Machine (UDO-SVM) methods. The correlation coefficients (R) of the best model predicted logK were 0.9332 for the training set and 0.9252 for the test set. This paper provided a novel method that only used chemical structures to develop efficient and stable models for testing and estimating association constants of the cesium chelates. This method will be an experimental guide to find macrocyclic compounds which are used to detect and minimize the radiocesium pollution.