caa a22 4500 445878819 CHVBK 20180317145535.0 cr unu---uuuuu 170323e20111101xx s 000 0 eng 10.1007/s11030-011-9325-2 doi (NATIONALLICENCE)springer-10.1007/s11030-011-9325-2 Theoretical study of GSK−3 α [Elektronische Daten] neural networks QSAR studies for the design of new inhibitors using 2D descriptors [Isela García, Yagamare Fall, Xerardo García-Mera, Francisco Prado-Prado] Glycogen synthase kinase-3 (GSK-3) targets encompass proteins implicated in AD and neurological disorders. The functions of GSK-3 and its implication in various human diseases have triggered an active search for potent and selective GSK-3 inhibitors. In this sense, QSAR could play an important role in studying these GSK-3 inhibitors. For this reason, we developed QSAR models for GSK−3α, linear discriminant analysis (LDA), and artificial neural networks (ANNs) from nearly 50,000 cases with more than 700 different GSK−3α inhibitors obtained from ChEMBL database server; in total we used more than 20,000 different molecules to develop the QSAR models. The model correctly classified 237 out of 275 active compounds (86.2%) and 14,870 out of 15,970 non-active compounds (93.2%) in the training series. The overall training performance was 93.0%. Validation of the model was carried out using an external predicting series. In these series, the model classified correctly 458 out of 549 (83.4%) compounds and 29,637 out of 31,927 non-active compounds (83.4%). The overall predictability performance was 92.7%. In this study, we propose three types of non-linear ANN as alternative to already existing models, such as LDA. Linear neural network: LNN: 236:236-1-1:1 which had an overall training performance of 96% proved to be the best model. In addition, we did a study of the different fragments of the molecules of the database to see which fragments had more influence in the activity. This can help design new inhibitors of GSK−3α. This study reports the attempts to calculate, within a unified framework probabilities of GSK−3α inhibitors against different molecules found in the literature. Springer Science+Business Media B.V., 2011 GSK−3 α nationallicence QSAR nationallicence Artificial neural network nationallicence Linear neural network nationallicence Fragment contribution nationallicence Linear discriminant analysis nationallicence García Isela Faculty of Chemistry, Department of Organic Chemistry, University of Vigo, Vigo, Spain aut Fall Yagamare Faculty of Chemistry, Department of Organic Chemistry, University of Vigo, Vigo, Spain aut García-Mera Xerardo Department of Organic Chemistry, University of Santiago de Compostela, 15782, Santiago de Compostela, Spain aut Prado-Prado Francisco Department of Organic Chemistry, University of Santiago de Compostela, 15782, Santiago de Compostela, Spain aut Molecular Diversity Springer Netherlands 15/4(2011-11-01), 947-955 1381-1991 15:4<947 2011 15 11030 https://doi.org/10.1007/s11030-011-9325-2 text/html Onlinezugriff via DOI 1 research-article jats NATIONALLICENCE 856 40 https://doi.org/10.1007/s11030-011-9325-2 text/html Onlinezugriff via DOI NATIONALLICENCE 700 1- García Isela Faculty of Chemistry, Department of Organic Chemistry, University of Vigo, Vigo, Spain aut NATIONALLICENCE 700 1- Fall Yagamare Faculty of Chemistry, Department of Organic Chemistry, University of Vigo, Vigo, Spain aut NATIONALLICENCE 700 1- García-Mera Xerardo Department of Organic Chemistry, University of Santiago de Compostela, 15782, Santiago de Compostela, Spain aut NATIONALLICENCE 700 1- Prado-Prado Francisco Department of Organic Chemistry, University of Santiago de Compostela, 15782, Santiago de Compostela, Spain aut NATIONALLICENCE 773 0- Molecular Diversity Springer Netherlands 15/4(2011-11-01), 947-955 1381-1991 15:4<947 2011 15 11030 Metadata rights reserved Springer special CC-BY-NC licence nationallicence BK010053 XK010053 XK010000 NATIONALLICENCE NATIONALLICENCE NL-springer