Flexible 3D alignment and its application in virtual screeningPresentation by Adrián Kalászi, et al · August 2009
Tackling the conformational flexibility of molecular structures is an innate challenge in most molecular modeling applications ranging from pharmacophore elucidation to virtual screening. Conformational sampling is the most widely used technique to alleviate the computational complexity of modeling flexible three dimensional molecules. Though computationally tractable, yet this approach has some drawbacks.
Full abstractMost importantly, it is prone to miss biologically relevant conformations. Representing flexible molecules on a continuous scale without the need of discrete sampling provides much higher degree of reliability and accuracy. But how can we address the complexity challenge and cope with the continuous flexibility within a manageable computational time frame?
A flexible 3D alignment method that overlays molecules by optimizing a potential function similar to the intersection of their molecular volumes has been developed. This method, based on the analytical representation of the conformational flexibility, is capable of aligning two or more chemical structures to very high accuracy. Nevertheless, the approach offers reasonable performance for drug-like molecules.
A generalization of the flexible conformational analysis apparatus enables the exploration of the entire conformational space of a molecule. The statistical analysis of this blurred spatial region provides a fairly low dimensional molecular descriptor which serves as the basis for a high throughput 3D virtual screening technique.
Download
Screening a Virtual Compound SpacePresentation by Szabolcs Csepregi, Ferenc Csizmadia, Szilárd Dóránt, Nóra Máté, György Pirok, Zsuzsanna Szabó, Jenő Varga, Miklós Vargyas · July 2004
From virtual screening to de novo drug design, there is a wide spectrum of tools available for computer aided drug discovery. The methods developed have their strengths as well as their weaknesses.
Full abstractOn one end of the spectrum virtual screening techniques, for instance, are fast and guarantee the availability of virtual hits for in vitro tests. However, the set of available chemicals is limited (10 000 000 is probably a good upper estimation for the number of known chemicals) and this implies a limited usability of virtual screening.
Towards the other end of the spectrum of discovery tools de novo drug design methods try to alleviate the problem of a finite compound space by suggesting novel molecular structures. Often, these are assembled by linking small molecular fragments in a procedure which is more or less chemically unaware. As a consequence the synthetic accesibility of de novo designed ligands is often a problem.
A method, which combines advantages of virtual screening with advantages of de novo design while eliminating their disadvantages has been implemented. This method deploys virtual screening over an ever growing space of virtual compounds that are likely to be synthetically accessible. To achieve this, commercially available chemical structures are combined by smart chemical reactions . These, unlike generic reaction equations that generate all possible products that are topologically relevant, are capable of eliminating products that are unlikely to be accessible via chemical synthesis.
In addition to synthetic accessibility there is another advantage of this method: no design is involved during structure generation. Consequently structures will be diverse both chemically and biologically and they can be screened for various targets simultaneously or at a later time so no potential lead is lost.
The applicability and efficiency of the method will be demonstrated with various examples.
Workshop on Chemical Information, July 2, Lausanne, Switzerland
Download
Combining fingerprints and other descriptors in virtual HTSPresentation by Miklós Vargyas, Zsuzsanna Szabó, Ferenc Csizmadia, György Pirok, Modest von Korff, Mathias Steger · April 2004
Choosing the right combination of the available descriptors is a tedious work when a virtual screening experiment is set up. Additionally, some descriptors may allow several parameters that increase the degrees of freedom dramatically. Finally, when comparing descriptor values one can choose from numerous dissimilarity metrics. To cope with this freedom of choice an automated optimization tool has been implemented.
Full abstractThis tool, which can be considered as an alternative to dimension reduction applied in QSAR, has proved to be successful in helping chemists to choose suitable descriptors, metrics and parameter values for virtual screening. It will be demonstrated that optimization can increase the hit ratio of the virtual screening procedure by 1 to 3 orders of magnitude.
227th ACS National Meeting · March 28 – April 1, 2004 · Anaheim, California
Download
Optimized Virtual ScreeningPresentation by Miklós Vargyas, Zsuzsanna Szabó, György Pirok, Ferenc Csizmadia, , Modest von Korff · November 2003
Drug research is often termed as searching for a needle in a haystack. Virtual screening is widely recognised as a valuable tool to effectively reduce the size of the ‘haystack’ by about one order of magnitude. In this presentation a technique that can further improve the efficiency of the screening procedure is proposed.
Full abstractWe focus on topological pharmacophore similarity based search where only a set of known active 2D structures is known. The pharmacophores of these structures are analysed by perceiving the pharmacophoric characteristic of each individual atom. Pharmacophore patterns are transformed into a topological cross correlation histogram. These correlation histograms are molecular descriptors that represent the pharmacophoric character of structures in a mathematically tractable form. Proximities (metrics) like the Euclidean distance and the Tanimoto coefficient are applied to estimate the dissimilarity between two such descriptors. The canonic formulae of the proximities are extended with weights and other parameters to help bias the metrics behaviour when comparing two compounds. Parameters are optimized in an automated training process that uses a subset of the target library and a subset of the known active structures. The optimized proximities are then passed on to an independent validation stage, which evaluates by calculating the enrichment ratio achieved within the virtual screening process.
Optimized virtual screening is capable of reducing the size of the ‘haystack’ by another order of magnitude (in some cases an even higher reduction is achieved) and it can also lead to scaffold hopping. The method is generic enough to adapt to other molecular descriptors and metrics. The efficiency of the method and cross-validated results will be presented.
eCheminformatics 2003, November 10-15, 2003
Download
Virtual screening using active set dependent optimization of dissimilarity metricsPresentation by Miklós Vargyas, Zsuzsanna Szabó, , Korff Modest, György Pirok, Ferenc Csizmadia · September 2003
The efficiency of virtual screening in drug discovery greatly depends on three factors: (1) pharmacophore point perception (2) representation of molecular structures with a descriptor, (3) dissimilarity metric to capture matching patterns in the descriptors. In this presentation methods tackling all three key factors will be discussed.
Full abstractPharmacophore point perception relying on the calculation of the protonation state of atoms and the partial charges at a user-defined pH assigns generalized types to atoms. Topological cross-correlation of these generalized atom types provides a compact representation of pharmacophores, however, the flexibility and shape of molecular structures is poorly represented. To overcome this problem, fuzzy smoothing of descriptors is introduced.
Virtual screening calculates the dissimilarity between a pair of descriptors using various metrics. The use of metrics comprising numerous tunable parameters set by an optimization procedure can lead to 250-fold enrichment over random.
Examples and further possible applications will be discussed.
226th ACS National Meeting, New York, USA, Sept 7-11, 2003.
Download
Virtual Screening Using Fingerprints Part II. Enhanced Search by Optimized Dissimilarity MetricsPresentation by Zsuzsanna Szabó, Miklós Vargyas, Modest von Korff, , Ferenc Csizmadia, György Pirok · August 2003
A large compound database is explored for structures that bare a similarity to a few given query structures, where the molecular similarity refers to the match of chemical, pharmacological and biological properties of two compounds.
Full abstractIn the starting part of the presentation the common marks, properties of two molecules are described. Molecular structures are encoded into molecular descriptors, e.g. fingerprints which can be handled numerically. That led to the method of virtual screening using these fingerprints. Fuzzy pharmacophore fingerprints offer a way to enhance the performance of screening and the introduction of tunable parameters in dissimilarity metrics can increase the performance too. In the second part of the presentation the optimalization of these parameters are discussed including the optimalization goal, the hypothesis, the examples of the hypothesis and the results. The pharmacophore perception is implemented within the ChemAxon’s JChem software tools.
Drug Discovery Technology, Boston, USA, August 10-15, 2003
Download
Virtual Screening Using Fingerprints Part I. A Hybrid Approach to Pharmacophore Point PerceptionPresentation by Miklós Vargyas, , Modest von Korff, György Pirok, Ferenc Csizmadia · August 2003
About 90% of drug targets cannot be crystallized. If the 3D structure of the active site is not available, a pharmacophore model is created. Such models are typically based on compounds that are known to bind to the target receptor. When creating pharmacophore models (either in 2D or in 3D) the pharmacophoric characteristic of known actives, as well as of candidates tested, have to be recognized.
Full abstractThe presentation gives an overview about the creation of pharmacophore models and perceptions. One possible approach to create pharmacophore models from 2D structures is to count frequencies of all atom-based pharmacophore point pairs with respect to their topological distance. On the other hand the rule based pharmacophore perception focuses on the molecules that are localized to moieties of the structure, defining whether the pharmacophore type of an atom will be a donor or an acceptor by rule. However, this perception has got two major problems: (1) large number of rules are needed to cope with exceptions; (2) different pH implies different rules. In the following section of the presentation the method of the hybrid approach for pharmacophore perception is thoroughly explained with its new rule-base and with results. The pharmacophore perception is implemented within ChemAxon’s JChem software tool.
Drug Discovery Technology, Boston, USA, August 10-15, 2003
Download
