Turning 3D-QSAR weakness into strength with Open3DALIGN & Open3DQSAR

The traditional 3D-QSAR workflow starts with an alignment of ligands in their putative bioactive conformation. The initial step is then followed by calculation of molecular interaction fields (MIFs), extraction of relevant information from the latter by statistical analysis and correlation with activity. The resulting model may have the power to predict the activity of new molecules before they are synthesised and tested [1]. The major weakness of 3D-QSAR methodologies is their dependency on the underlying alignment. Even when the bioactive conformation of a template molecule is known, usually from an experimentally determined structure of a ligand-target complex, the alignment procedure itself is a difficult and time-consuming operation, especially in the presence of flexible or structurally heterogeneous ligands. When the structure or even the identity of the target is not known, it becomes very difficult to hypothesize a univocal and reliable alignment. Unfortunately, the lack of knowledge of the target’s structure is also the situation where a ligand-based approach would be most desirable, since it becomes basically the only option for computer-aided drug design. Herein we present Open3DALIGN [2], an open-source tool capable of running conformational searches (by a TINKER [3]-based QMD engine) and generating unsupervised ligand alignments. The initial alignment bottleneck in 3D-QSAR may be overcome by ranking a large number of possible alignments on the basis of their consistency and the predictive performance of the corresponding 3D-QSAR models, built and evaluated with Open3DQSAR [4]. This procedure allows to formulate unbiased hypotheses on the bioactive conformation of a series of ligands in the absence of prior knowledge of the target’s structure or ligand’s SAR. Therefore, the 3D-QSAR alignment dependency is turned from a weakness into strength: binding mode hypotheses may be challenged according to their ability to explain experimental activities from a purely ligand-based perspective. Promising results were obtained applying this methodology on eight benchmark literature datasets [5]. Most recent developments of Open3DALIGN are discussed, namely the implementation of a novel, all-atom alignment algorithm in addition to the pharmacophore-based one relying on Pharao [6]. 1. Cross, S.; Cruciani, G. Molecular fields in drug discovery: getting old or reaching maturity? Drug Disc. Today 2010, 15, 23-32. 2. Tosco, P.; Balle, T. Open3DALIGN: an open-source software aimed at unsupervised molecular alignment; http://open3dalign.org (accessed Jan 31, 2011). 3. TINKER - Software Tools for Molecular Design, version 5.1; http://dasher.wustl.edu/tinker/ (accessed Jan 31, 2011). 4. Tosco, P.; Balle, T. Open3DQSAR: an open-source software aimed at high-throughput chemometric analysis of molecular interaction fields; http://open3dqsar.org (accessed Jan 31, 2011). 5. Tosco, P.; Balle, T. Poster communication at the 6th German Conference on Chemoinformatics, Goslar, Germany, 7-9 November 2010; http://va.gdch.de/wwwdata/abstracts/5412/5412_0066.pdf (accessed Jan 31, 2011). 6. Pharao version 3.0.3; http://www.silicos.be/pharao.html (accessed Jan 31, 2011).