|
Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT
| ABSTRACT - Internet Electron. J. Mol. Des. September 2005, Volume 4, Number 9, 625-646 |
General and Independent Approaches to Predict HERG Affinity Values
Elena Fioravanzo, Nicola Cazzolla, Lucia Durando, Cristina Ferrari, Massimo Mabilia, Rosella Ombrato, and Marco Daniele Parenti
Internet Electron. J. Mol. Des. 2005, 4, 625-646
|
Abstract:
The protein product of the human ether-a-go-go gene (hERG) is
a potassium channel that when inhibited may lead to cardiac
arrhythmia. At present, various in vivo and in vitro models for
QT prolongation and subsequent arrhythmia exist but they may
not be entirely predictive for humans. Consequently, a fast and
reliable in silico model to assess hERG affinity values would
increase the screening rate and would also lower the cost
compared to experimental assay methods. In this communication
different approaches were employed to predict hERG K+ channel
affinities. First of all, different QSAR models were developed
employing various molecular descriptors. Then, independent
software were used to predict hERG activity values: Qikprop and
PASS. The software QikProp (Schrödinger, L.L.C) allows to
predict pharmaceutically relevant properties for organic
molecules, starting from their 3D structures and employing
calculated physically significant descriptors. In addition to cell
permeability, logP, solubility, blood/brain barrier permeability,
the program can also predict hERG K+ channel affinity values.
As an independent approach, the program PASS PRO -
Prediction of Activity Spectra for Substances - (V. Poroikov, D.
Filimonov & Associates) that can predict several hundreds
biological activity probability values, such as pharmacological
effects, mechanisms of action, toxicity and metabolism reactions,
was trained to predict the probability of hERG activity. The
availability of different and independent methods and models
able to predict hERG activity allows the application of a
consensus criterion to be used as a filter in the discovery process.
Five QSAR models were obtained with Q2 values ranging from
0.65 to 0.98 and SDEP values ranging from 1.2 to 0.9.
Employing together QikProp, PASS and QSAR predictions, we
obtained a consensus criterion that applied to 67 molecules yields
a Matthews correlation of MCC = 0.71, 5 FP and 3 FN. In the
light of such result, our consensus score can be used as a
powerful in silico screening for drug discovery processes.
|