Identification of Dopamine D1 Receptor Agonists and Antagonists under Existing Noise Compounds by TFS-based ANN and SVM

Abstract

This paper describes classification and prediction for pharmacologically active classes of drugs under the presence of noise chemical compounds. Dopamine D1 receptor agonists (63 compounds), antagonists (169 compounds) and other drugs (696 compounds) were used for the work. Each drug molecule was characterized with Topological Fragment Spectra (TFS) reported by the authors. TFS-based artificial neural network (TFS/ANN) and support vector machine (TFS/SVM) were employed and evaluated for their classification and prediction abilities. It was concluded that the TFS/SVM works better than TFS/ANN in both the training and the prediction.

Identification of the Dual Action Antihypertensive Drugs Using TFS-Based Support Vector Machines

“…SVM was successfully applied to the classification of biological activity of drugs [15] [16]. In addition, SVM deterministically classify the instance being "active" or "inactive" on the basis of the discriminant surface that was determined in advance.…”

“…In addition, SVM deterministically classify the instance being "active" or "inactive" on the basis of the discriminant surface that was determined in advance. In the previous studies [15][17], we investigated the classification of dopamine agonists and antagonists, in which each drug belongs to a single class. It was shown that SVM performed better than the artificial neural networks with multiple output neurons [15], and the SVM in conjunction with topological fragment spectra (TFS) descriptors, we called TFS-based SVM, could successfully learn the relationships between chemical structures of the drugs and their biological activity.…”

“…In the previous studies [15][17], we investigated the classification of dopamine agonists and antagonists, in which each drug belongs to a single class. It was shown that SVM performed better than the artificial neural networks with multiple output neurons [15], and the SVM in conjunction with topological fragment spectra (TFS) descriptors, we called TFS-based SVM, could successfully learn the relationships between chemical structures of the drugs and their biological activity. The authors also reported that the TFS-based SVM approach applied to biological activity profiling of drugs including multi-label compounds [18].…”

Predictive Activity Profiling of Drugs by Topological-Fragment-Spectra-Based Support Vector Machines

“…Nevertheless, the present approach worked very well. In another work, 29 we investigated TFSbased machine learning using unbalanced data sets. The results suggested that training by TFS-based ANN strongly depends on the size of the data set for each class.…”

“…In a previous work, we reported that an ANN approach combined with the topological fragment spectra (TFS) allowed us to successfully classify dopamine antagonists that interact with four different types of dopamine receptors, and it could be applied to the prediction of activity for class-unknown compounds. It was also shown that SVM works for this type of problem – much better. However, in those cases, each compound belonged exactly to one single class, and no multilabeled compounds were included for the data sets.…”