| Abstract
| - A benchmark of several popular methods, Associative Neural Networks (ANN), Support Vector Machines(SVM), k Nearest Neighbors (kNN), Maximal Margin Linear Programming (MMLP), Radial Basis FunctionNeural Network (RBFNN), and Multiple Linear Regression (MLR), is reported for quantitative−structureproperty relationships (QSPR) of stability constants logK1 for the 1:1 (M:L) and logβ2 for 1:2 complexes ofmetal cations Ag+ and Eu3+ with diverse sets of organic molecules in water at 298 K and ionic strength 0.1M. The methods were tested on three types of descriptors: molecular descriptors including E-state values,counts of atoms determined for E-state atom types, and substructural molecular fragments (SMF). Comparisonof the models was performed using a 5-fold external cross-validation procedure. Robust statistical tests(bootstrap and Kolmogorov-Smirnov statistics) were employed to evaluate the significance of calculatedmodels. The Wilcoxon signed-rank test was used to compare the performance of methods. Individualstructure−complexation property models obtained with nonlinear methods demonstrated a significantly betterperformance than the models built using multilinear regression analysis (MLRA). However, the averagingof several MLRA models based on SMF descriptors provided as good of a prediction as the most efficientnonlinear techniques. Support Vector Machines and Associative Neural Networks contributed in the largestnumber of significant models. Models based on fragments (SMF descriptors and E-state counts) had higherprediction ability than those based on E-state indices. The use of SMF descriptors and E-state counts providedsimilar results, whereas E-state indices lead to less significant models. The current study illustrates thedifficulties of quantitative comparison of different methods: conclusions based only on one data set withoutappropriate statistical tests could be wrong.
|
