Generative Topographic Mapping-Based Classification Models and Their Applicability Domain: Application to the Biopharmaceutics Drug Disposition Classification System (BDDCS)

Abstract

Earlier (Kireeva et al. Mol. Inf. 2012, 31, 301-312), we demonstrated that generative topographic mapping (GTM) can be efficiently used both for data visualization and building of classification models in the initial D-dimensional space of molecular descriptors. Here, we describe the modeling in two-dimensional latent space for the four classes of the BioPharmaceutics Drug Disposition Classification System (BDDCS) involving VolSurf descriptors. Three new definitions of the applicability domain (AD) of models have been suggested: one class-independent AD which considers the GTM likelihood and two class-dependent ADs considering respectively, either the predominant class in a given node of the map or informational entropy. The class entropy AD was found to be the most efficient for the BDDCS modeling. The predominant class AD can be directly visualized on GTM maps, which helps the interpretation of the model.

Scaffold Diversity of Fungal Metabolites

“…The main disadvantage of PCA is that it is a linear mapping technique and is unable to map non-linear data. GTM is a nonlinear method that trains a Radial Basis Function (RBF) neuronal network to produce a mapping from an n-dimensional data space to a two dimensional latent space (Owen et al, 2011; Gaspar et al, 2013). For further explanation on each model, the reader is referred to the cited papers (Gaspar et al, 2013, 2015).…”

“…GTM is a nonlinear method that trains a Radial Basis Function (RBF) neuronal network to produce a mapping from an n-dimensional data space to a two dimensional latent space (Owen et al, 2011; Gaspar et al, 2013). For further explanation on each model, the reader is referred to the cited papers (Gaspar et al, 2013, 2015). To represent the chemical space using molecular fingerprints, a fingerprint array was assembled from the MACCS key fingerprint results, consisting of 166 bits in which each element is either 0 or 1 to indicate the absence or presence, respectively, of structural elements in the corresponding molecular structure.…”

Generative Topographic Mapping of Conformational Space

“…Part of this chromosome contains the GTM setup: number of nodes, number of RBF functions defining the manifold and their width, the regularization coefficient. [10,18,19] Another key parameter encoded by the chromosome is the type of descriptors to use -here, both already available, classical and on-purpose designed conformational descriptors that will be detailed below.…”

“…Like above-mentioned clustering techniques, its outcome may be highly method-dependent, and there is a wealth of dimensionality reduction algorithms -linear Principal Component Analysis (PCA), [5] and non-linear approaches -Self-Organizing Maps [6] (SOM), Multidimensional Scaling [7] (MDS), Stochastic Embedding, [8] 2D scaling with rubber bands, [9] and, in particular, Generative Topographic Maps [10] (GTM).…”

Progress in visual representations of chemical space

“…Newer applications include the analysis of property distribution and landscapes in the datasets [60], target and pharmacophore mapping of the natural product space [63], prediction of melting points for ionic liquids [64], classification of drugs according to the solubility and metabolism [65]. Nevertheless, in certain cases no significant difference can be found between the SOM and GTM [66].…”