Classification Methods

While QSAR methods aim to predict absolute compound activity, classification methods attempt to bin compounds by their potential hERG inhibition. The earliest example of a hERG-based classification was reported by Roche and coworkers.57 A total of 244 compounds representing the extremes of the data set (<1 and>10mM for actives and inactives, respectively) were modeled with a variety of techniques such as substructure analysis, self-organizing maps, partial least squares, and supervised neural networks. The descriptors chosen included pKa, Ghose-Crippen,58 TSAR,59 CATS,60 Volsurf,61 and Dragon62 descriptors. The most accurate classification was based on an artificial neural network. In the validation set containing 95 compounds (57 in-house and 38 literature IC50 values) 93% of inactives and 71% of actives were predicted correctly.

In a decision tree-based approach to constructing a hERG model using calculated physicochemical descriptors, Buyck and co-workers56 used three descriptors - ClogP, calculated molar refractivity (CMR), and the pKa of the most basic nitrogen - to identify hERG blockers within an in-house data set. With IC50 = 130 nM as a cutoff, factors suggestive of hERG activity were determined to be ClogPX 3.7, 110 pCMR <176, and pKa maxX7.3.

A combined 2D/3D procedure for identification of hERG blockers was proposed by Aronov and Goldman.63 A 2D topological similarity screen utilizing atom pair64 descriptors and an amalgamated similarity metric termed TOPO was combined with a 3D pharmacophore ensemble procedure in a 'veto' format to provide a single binary hERG classification model. A molecule flagged by either component of the method was considered a hERG active. In the course of 50-fold cross-validation of the model on a literature data set containing 85 actives (threshold HERG IC50 = 40 mM) and 329 inactives, 71% of hERG actives and 85% of hERG inactives were correctly identified. The model utilizing the TOPO metric was shown to be superior to a number of other 2D models using the receiver operating characteristic metric. Additionally, five of eight (62.5%) hERG blockers were identified correctly in a 15 compound in-house validation set. Most of the statistically significant pharmacophores from the ensemble procedure were three-feature [aromatic]-[positive charge]-[hydrophobe] combinations (Figure 8a-b) similar to those reported by Cavalli and colleagues44; however, a novel three-point pharmacophore containing a hydrogen bond acceptor was also proposed (Figure 8c). The presence in hERG blockers of the acceptor functionality pointing toward the selectivity pore agrees with the previous observations15,40,44 of a potential for polar interactions with the side chains of Thr623 and Ser624 to stabilize the hERG-ligand complex.

Testai and co-workers65 evaluated a set of 17 antipsychotic drugs, all of them associated with reports of torsadogenic cardiotoxicity. The search for a common molecular feature that may be a requirement for hERG blockers focused on measuring several different distances between atoms that could constitute a hERG-active template. The authors hypothesized that such a template for hERG-active ligands consists of a hydrocarbon chain, three or four atoms long, serving as a spacer between a basic sterically hindered nitrogen atom, and a second, more variable, moiety. Focusing on the distance between the basic nitrogen and this second moiety, Testai and co-workers observed that for all of the compounds in the data set the distance converged in the range between 4.32 and 5.50A (average = 4.87 A).

Figure 8 Three three-point pharmacophores (a-c) for in silico hERG block prediction (from Aronov, A. M.; Goldman, B. B. Bioorg. Med. Chem. 2004, 12, 2307-2315). Colored features correspond to positive charge (blue), hydrogen bond acceptor (red), and hydrophobic (green). (Reproduced with permission from Aronov, A. M. Drug Disc. Today 2005, 10, 149-155 © Elsevier.)

Figure 8 Three three-point pharmacophores (a-c) for in silico hERG block prediction (from Aronov, A. M.; Goldman, B. B. Bioorg. Med. Chem. 2004, 12, 2307-2315). Colored features correspond to positive charge (blue), hydrogen bond acceptor (red), and hydrophobic (green). (Reproduced with permission from Aronov, A. M. Drug Disc. Today 2005, 10, 149-155 © Elsevier.)

H2N

Was this article helpful?

0 0
Lower Your Cholesterol In Just 33 Days

Lower Your Cholesterol In Just 33 Days

Discover secrets, myths, truths, lies and strategies for dealing effectively with cholesterol, now and forever! Uncover techniques, remedies and alternative for lowering your cholesterol quickly and significantly in just ONE MONTH! Find insights into the screenings, meanings and numbers involved in lowering cholesterol and the implications, consideration it has for your lifestyle and future!

Get My Free Ebook


Post a comment