Premium
An insight QSPR‐based prediction model for stability constants of metal‐thiosemicarbazone complexes using MLR and ANN methods
Author(s) -
Quang Nguyen Minh,
Nhung Nguyen Thi Ai,
Van Tat Pham
Publication year - 2019
Publication title -
vietnam journal of chemistry
Language(s) - English
Resource type - Journals
eISSN - 2572-8288
pISSN - 0866-7144
DOI - 10.1002/vjch.201900070
Subject(s) - quantitative structure–activity relationship , semicarbazone , stability (learning theory) , computational chemistry , chemistry , metal , organic chemistry , stereochemistry , computer science , machine learning
In the present investigation, the stability constants (log β 12 ) of complexes (ML 2 ) between metal ions (M) and thiosemicarbazones (L) were used as an endpoint in the quantitative structure‐property relationship (QSPR) approaches. The molecular descriptors of the experimental complexes were calculated from the conformation with the lowest binding free energy by means of semi‐empirical PM7 method. QSPR models were developed by using multivariate linear regression (MLR) and artificial neural network methods (ANN). The best QSPR models found out three important descriptors as knotp, Cosmo Area and Hmin in the metal‐thiosemicarbazones complexation. The final QSPR MLR model had shown satisfactory statistical performance; training (R 2 train ) and prediction (Q 2 LOO ) determination coefficient of 0.9274 and 0.8784, respectively. Meanwhile, the statistical results of QSPR ANN model received the value of 0.9844 and 0.9898. The models also ratified strict statistical validation tests (Q 2 test ) for external predictivity with the QSPR MLR and QSPR ANN value of 0.8321 and 0.8953, respectively. A series of new metal‐thiosemicarbazones complexes were designed based on the descriptor of the models and predicted the stability constants of the complexes.