MOE is a software system designed to support Cheminformatics, Molecular Modelling, Bioinformatics, Virtual Screening, Structure-based-design and can be used to build new applications based on SVL (Scientific Vector Language). Given the breadth of uses this review can only highlight a small part of the potential scope.
The main window of MOE shows the default interface, the top menu bar provides access to a range of applications and tools, below the top menu the is a commandline interface that can be used to input SVL commands. The buttons to the right of the rendering area are short cuts to commands that are often buried several levels down in the top menu system, these are also available by pressing the alt key.. These include many of the rendering and labelling options, angle and distance measurements, and access to the molecular builder. At the foot of the window are dials used for rotating, translating and zooming. These actions can also be accessed using the mouse by pressing combinations of the alt, ctrl and shift keys, or using the middle button of a tree-button mouse.
Perhaps one of the most unusual features of MOE is the ability to customise the interface using SVL, either to introduce custom features or to modify the interface for particular users e.g. novice users or protein modellers as shown below.
To import a structure use the File:Open Menu, MOE can read most common file formats (sdf, SMILES, pdb, mol2) as well as the internal .moe file type. A variety of structure builders are also supplied that can be used to build or edit systems of varying complexity, from small molecules and carbohydrates to proteins and crystals. The molecule builder is limited but functional but does allow the user to enter structures or fragments as SMILES strings.
Once built the structure can be minimised using a variety of forcefields these include MMFF94, AMBER, CHARMM and semi-empirical energy functions (PM3, AM1, MNDO). Conformational analysis using either a systematic or a stochastic search using random rotation of bonds is available. The generated conformations can be stored in a database and analyzed later.
The moe molecular databases also provide access to a variety of other tools including sorting and structure based searching and the calculation of a vast range of molecular descriptors. The plot functionality can be used to provide a simple visual comparison of the properties.
Alternatively the calculated descriptors can then be used in a principle component analysis, the results can be displayed as a 3D graph in the rendering window. Clicking a data point on the graph highlights the molecule in the database. The data can also be used to construct a QSAR model including cross-validation and activity prediction. Tools are then available for screening real or virtual libraries to select likely novel active compounds using either binary tree or linear regression models.
The virtual libraries can be generated using QuaSAR-CombiGen, this generates a fully-enumerated combinatorial library from a scaffold database and a set of substituent R-group databases. This can be pruned using clustering tools to provide a diverse sub-set of molecules that represent the molecular space covered by the library. MOE also provides a comprehensive set of Structure-Based design tools, including active site finder, a MultiFragment Search (MFS) to help understand the interactions between chemical functional groups with the active site of a receptor, protein surface calculations and of course molecular docking used to search for favorable binding configurations between one or more small, flexible ligands and a protein target. This is achieved by using of poses generated from the pool of ligand conformations (these can be generated previously and stored in a database). The Dock application provides a framework for the integration of multiple placement methodologies; each such placement methodology will have different properties. Optionally, the generated poses are constrained to satisfy an arbitrary pharmacophore query. Such a query is used to bias the search towards known important interactions. Each pose generated by the placement methodology is subjected to scoring in an effort to identify the most favorable poses. Typically, scoring functions emphasize favorable hydrophobic, ionic and hydrogen bond contacts. MOE also includes a set of tools for protein modelling, including homology searching, secondary structure prediction, sequence and structure alignment and tools for building homolgy models. As mentioned before MOE is written in SVL a “chemically aware” vector language ideally suited to problems involving computations over large amounts of data. By manipulating data in vector form, an operation can be performed over an entire data set using a single instruction. A detailed tutorial on SVL is available and an extensive range of example code are available for users (http://svl.chemcomp.com/).
