Toolkit for 2D and 3D molecular screening
Ligand-based screening of large molecular databases has become an invaluable tool in early phase drug-discovery. The Screen Suite is a ligand-based high throughput virtual screening package that provides powerful tools for chemical (2D) and shape (3D) similarity searches of large molecular libraries. Screen provides several different fingerprints sets, such as the ChemAxon chemical fingerprint, pharmacophore fingerprint and ECFP/FCFP as well as several dissimilarity metrics and metrics optimization that can be used to tune your processes to achieve optimal search results.
The Screen Suite is available commercially as 2D Descriptor package, Screen2D (includes the ChemAxon Chemical Fingerprint) and Screen3D. All components of the Screen Suite are accessible as command line applications, and also include a full Java and .NET based Application Programming Interface (API) through which they can be integrated into existing in-house or commercial third party applications.
Descriptor package in Screen
Flexible descriptor accessibilityThe Screen2D Descriptor package contains fingerprint generators for chemical structures, pharmacophores, extended connectivity (ECFP) and functional class (FCFP) descriptors and can be used to generate and tune fingerprints for use in any screening process. Descriptor generation is also supported through ChemAxon’s Chemical Terms scripting language, to let create complex fingerprints which consider single or combined molecular attributes that can be based on topological or physicochemical properties e.g. partition coefficients, hydrogen bonding donor-acceptor accessibility or acidic properties.
Diverse fingerprints for different purposes
The BCUT descriptors are based on a special matrix representation of the connectivity table of the molecules extended with certain physical and physicochemical atomic properties and are capable of encoding compound properties relevant to intermolecular interactions.
ChemAxon’s BCUT descriptor encodes atomic charge, polarizability and hydrogen bonding donor-acceptor properties. The BCUT descriptors are widely used in diversity analyses and QSAR applications.
The chemical fingerprint is a path-based bit-string descriptor containing certain limited information on the molecular structure. The chemical fingerprint can be used in database handling for structural searches (e.g. substructure and full structure searching) and for similarity searches. It also can be used for the diversity analysis of compound libraries.
The pharmacophore fingerprints are atom-based fingerprints designed to characterize the compounds’ binding related structural and physicochemical properties. Pharmacophoric features include all those binding related structural or physicochemical properties of chemical compounds that are thought to be responsible for pharmacological activity.
Chemical attributes taken into account usually include hydrogen bond donor/acceptor capability, charge, hydrophobicity and aromaticity. The fingerprint generation is based on simple statistics determined by the relative arrangement according to topological distances of pharmacophore points, which results in the two dimensional nature of the fingerprint.
The Screen 5.4 Suite supports the use of circular fingerprints in the screening process. In contrast to path-based fingerprints ECFP/FCFP are not suitable for substructure search, but are relevant for full structure and similarity searches. For similarity searches circular fingerprints are known to yield more consistent results than path-based descriptors.
The use of the FCFP gives researchers the freedom to include arbitrary physical, physicochemical or any arbitrary data in the fingerprint itself. Calculation of these properties can be supported by the Chemical Terms language.
Screen2D is a ligand-based virtual high throughput screening platform, that can be used to carry out similarity searches on large molecular databases using different descriptors and several different metrics to quantify dissimilarity of compounds. Screen functionality is available through API and command line, via Pipeline Pilot and KNIME connectors, and some features are implemented within Instant JChem and JChem for Excel desktop applications.
The screening process is fast, with up to 100,000 compounds per second being screened on a single desktop PC.
Great flexibility in descriptor usage
Screen’s 2D descriptor package includes many descriptors, such as BCUT, Chemical and Pharmacophore fingerprints and ECFP/FCFP. Besides the extendibility of the FCFP by user defined additional parameters, researchers can use arbitrary sets of scalars (i.e. physicochemical properties) as descriptors or any in-house generated fingerprints. Screen’s modularity allows for the user to use their own well established and validated descriptor sets.
Screen2D comes with a set of metrics to measure dissimilarity between compounds. It supports the Euclidean, Tanimoto and Tversky metrics and also modified versions, such as the Dice metrics. Screen2D also supports metrics optimization – by selecting appropriate training sets, the optimization step increases enrichment significantly, giving smaller, more focused hit sets.
In addition to topology based screening, the Screen Suite includes 3D shape similarity based searching using Screen3D. Screen3D introduces a novel flexible alignment technique, which provides reasonable alternative hit results as compared to previous 3D screen engines. Screen3D is also available both through the API and as a command line program.
High speed 3D shape similarity searching
A key feature of Screen3D is its high speed. Currently Screen3D is ten times faster than any alternative approach which, besides the novel alignment algorithm is achieved by introducing a preprocessing step carried out separately from the actual screening process. Screen3D also compares very well in providing better enrichment in the first 1% of the hits than most alternative software. Screen3D’s scalability makes it suitable to search in very large molecular databases.
Multiple 3D screening strategies
Screen3D can be used to carry out different shape similarity searches. Beyond fully flexible matching Screen3D can align flexible database compounds to rigid query molecules (rigid-flexible), or alternatively to align rigid database compounds to rigid query molecules (rigid-rigid) using on-the-fly generated conformation ensembles.
Two types of 3D similarity screening methods are available in Screen3D: “Shape” and “Match”. Due to their dissimilar methodologies these approaches capture different aspects of binding. While “Shape” evaluates full molecular shape similarity measures that can model ligand binding to a protein’s active site, “Match” intends to explore key atoms in ligand binding while the actual shape of the molecule is not considered.
2D structures for 3D searches
Screen3D does not need 3D input coordinates to carry out 3D shape similarity comparison on structural libraries. Databases and query structures containing only 2D coordinates can be applied; the 3D generation is carried out on-the-fly during the screening process.
Hits can be easily visualized
Screen3D is capable of generating outputs that can be directly loaded to MarvinView or MarvinSpace to visualize 3D shape similarity comparison hits along with the aligned compound structures and similarity scores. 3D alignment can also be carried out in Marvin Space against any rigid conformer of a target compound – query structures can be aligned directly to substrate structures as present in a protein’s active site.