in silico Chemogenomics Applications in Drug Discovery

Analysing chemogenomic data is a never-ending learning process as the completion of a huge matrix is sought. In this matrix, ligands/drugs are annotated with experimentally determined binding affinities of protein–ligand complexes and multidimensional biological phenotypes that reflect biological networks and polypharmacology.

The prediction of targets and off-target binding by in silico chemogenomics leads to challenging predictions: adverse reactions, false-positives in cell-based reporter gene assays and alternative mechanisms of action. The ligand-target space is mostly described by 1-D or 2-D descriptors, though 3-D descriptors are becoming more attractive with the growing number and diversity of available target/ligand complexes.

On 15 October 2008 we will hold an eCheminfo Community of Practice conference session at Bryn Mawr College to be chaired by Fabrice Moriaud (MEDIT), where we have invited several leading scientists to share their experiences and research in the emerging field of in silico chemogenomics.

A description of the session with presentation abstracts follows:

In Silico-based chemogenomics

http://echeminfo.com/COMTY_confprog08chemogenomics

(Please follow continuation here to read abstracts.  Comments can be made at the end.)

Abstracts

Chemogenomic Space Exploration: Use of MED-SUMO Tools to Analyse Target and Ligand Space of the Kinome

Charles Andrianjara, Laboratoires Pierre Fabre, France

In recent years, growing interest has been shown toward the implementation of chemogenomic technology in the drug design process. This approach is recognized to be an alternative to the traditional “one-drug one-target” design strategy. The success of this new drug design paradigm is related to the degree of knowledge of the chemogenomic space, defined by the merging of target and ligand space.

Med-SuMo, a target based drug design tool, provides a new approach for characterising the target space, based on the local shapes and 3D chemical features of protein binding sites. This technology offers a new perspective of the target space, which is defined by the ligand binding site features instead of the protein primary sequences. Furthermore, the MED-SuMo descriptors are useful not only for similarity and dissimilarity analyses between protein binding sites from any 3D structure database (PDB, Corporate), but also for the identification of key residues involved in ligand affinity and selectivity.

In an attempt to retrospectively evaluate the impact of the chemogenomic approach in the design of kinase inhibitors, the MED-SuMo technology was used to characterize the target space constituted by the kinome family members. The first step of this study was thus focused on cluster analysis of protein kinases available in the PDB, by using the MED-SuMo descriptors. The aim of the previous step was to obtain more insight on the target space features. Next, the ligand space was constructed based on the structural and biological data extracted from the kinase inhibitor database of AurSCOPE. Finally, the target space was merged with the ligand space to validate the kinase classification methods and also to obtain kinase inhibitor SAR profiles. The inhibitor SAR and target information revealed by the merging of ligand and target space of the kinome will be discussed in this presentation. In addition, a comparative study of kinase classifications from two approaches: Med-SuMo binding site descriptors and protein primary sequence will be presented.

Drug Repurposing and Side-effect Elucidation by Statistical Chemical Similarity

Michael Keiser, UCSF

Chemically similar drugs often bind biologically diverse targets, yet many marketed drugs have been presumed selective for their intended targets at therapeutic concentrations. In this work, we use the Similarity Ensemble Approach (SEA) to uncover new chemical similarities of known drugs compared against a panel of 65,000 ligands organized into hundreds of target sets. Novel off-target links emerged, including the predictions that fluoxetine (Prozac), domperidone (Motilium), and tetrabenazine (Nitoman) may antagonize the beta-adrenergic, alpha1-adrenergic, and alpha2-adrenergic receptors, respectively. In addition, fluanisone and dimetholizine were both predicted to antagonize the alpha1-adrenergic and the 5-HT1A receptors. All of these prospective predictions were confirmed by experiment at nanomolar affinities. Relating drugs to receptor ligands by shared chemical patterns reveals the unexpected polypharmacology of existing drugs.

Doing more than just the Structure – Insights into Kinase Structure and Drug Discovery by Structural Genomics

Brian Marsden, Structural Genomics Consortium Oxford, UK

Structural genomics (SG) has significantly increased the number of novel protein structures of targets with medical relevance. In the protein kinase area, SG has contributed more than 50% of all novel kinases structures during the past four years and determined more than 35 novel catalytic domain structures. This has generated a wealth of freely-available reagents, assays, and inhibitor screening data which may be used for early drug development both in academic labs and in industry.

Kinase structures determined by SG provide enhanced opportunities for structural comparison within sub-families in order to expose novel structural motifs and mechanisms that may play a role in catalytic activity and activation. For example, we have identified novel domain-swaps of the activation loop in a number of ser/thr kinases which may provide a mechanism of auto-activation. Analysis of class II PAK kinase domains in various bound/un-bound states has elucidated a network of interactions around the back of the ATP-binding site that may transmit the state of the binding site to the catalytic area.

Many kinase structures include co-crystallised small molecules which in some cases have exposed novel chemotypes and binding modes. These also include high-resolution structures with well-known inhibitors bound to kinases which they were not originally designed for, giving additional insight into off-pathway interactions.

Despite the increased rate of protein kinase structure determination, there is much work still to be done. An analysis of the structural coverage of many kinases families reveals that there are significant portions which are under-represented. Similarly, inhibitor complexes with diverse inhibitors are only available for a few kinases. By choosing appropriate representative targets and probing with diverse compounds, the rate at which these gaps are closed can be significantly improved. SG will play a key role in this regard, as generated reagents continues to provide data beyond protein structure determination by identifying chemical probes, determining their binding modes and target specificity.

Cheminformatics Analysis of Polypharmacological Data

Alexander Tropsha, The Laboratory for Molecular Modeling, Division of Medicinal Chemistry and Natural Products, School of Pharmacy, UNC-Chapel Hill, Chapel Hill, NC 27599, USA

Modern experimental drug discovery efforts are increasingly focusing on the development of multi-target directed ligands that produce the desired pharmacological and pharmaceutical effects. Recent advances in high-throughput screening and multi-target testing of compound libraries coupled with the establishing of publicly available databases of biologically tested compounds call for the development of sophisticated computational tools and models of complex chemical genomics data. We define a dataset as complex if multiple measures of biological activity/property are reported for all (or most of) compounds in the entire chemical library. The examples of complex datasets include Pubchem, PDSP, DSS-Tox, and others. We shall consider emerging methodologies for analyzing complex chemogenomics datasets such as subspace clustering, database graph analysis, and others. We shall present models that relate compound structure to their multi-target profiles (as opposed to more traditional single target specific models). Modeling of complex chemogenomics databases present new challenges and new frontiers in molecular modeling.

Barry Hardy

eCheminfo Community of Practice

Chemogenomics eCheminfo cheminformatics chemoinformatics bioinformatics Medicinal Chemistry Computational Chemistry Virtual Screening Molecular Modelling Molecular Modeling pharmaceutical pharma meeting workshop training Oxford Critical Path toxicology Bursary Life Sciences Pharma Drug Discovery Research and Development Drug Development Healthcare Innovation Knowledge Management events

Comments

[Compounding Pharmacist]

Compound Pharmacy has gained much popularity in the field of medicine. The medications are equally effective and safe for sick patients who cannot take the actual medications due to their personal allergies.