We provide three different experiment types as part of our Virtual Laboratory Suite:
1) Support Vector Machine
2) Molecular Docking
3) Functional Properties
Support Vector Machines
A Support Vector Machine (SVM) is a machine learning technique designed to classify data. In this case we use SVMs, trained on datasets of known binding interactions, to predict whether your ligand of interest is a good candidate for binding with the protein(s) you are interested in. We have designed SVMs (for example, see our recent publication on using SVMs to predict binding interactions of P-gp with small molecules in PLoS ONE1 or Human Serum Albumin (HSA) with a variety of ligands in Bioinformatics2). In addition to our own SVMs, we implement SVMs published in peer-reviewed scientific articles to broaden the array of possible protein-ligand interactions we can predict.
In general, SVM techniques, developed by computational chemists, are not readily accessible to experimental scientists. Our aim is to provide a user-friendly interface, using rigorously tested computational methods to experimenters in order to aid in rational drug and experimental design.
As well as knowledge-based SVM predictions from published classifiers, we also offer de novo molecular docking experiments with AutoDock and MOPAC2007, as with our industry-leading product DockingServer.
Functional properties predicted include molecular weight, mean atomic polarizability, Randić’s molecular connectivity index, the Ghose–Viswanadhan–Wendoloski anti-inflammatory-like index, and the Lipinski rule among others. These property checks, whether on protein-ligand complexes or just on ligands allows for the approximate indications about potential bio-activity and suitability of a target molecule as a drug lead.
We provide structures and experiments geared towards ligand-binding predictions of several different proteins and protein groups. We are constantly expanding our repertoire of PDB files with an emphasis on those important for disease and drug design.