Optibrium

Matt presented this poster at ISSX in October 2011.

Abstract:

Many computational methods have been developed that predict the regioselectivity of metabolism by drug metabolising isoforms of the Cytochrome P450 class of enzymes (P450) [1-5]. Here we describe recent developments to a method for predicting P450 metabolism that combines quantum mechanical (QM) simulations to estimate the reactivity of potential sites of metabolism on a compound with a ligand-based approach to account for the effects of orientation and steric constraints due to the binding pockets of different P450 isoforms.

Matt presented this poster at ISSX in October 2011.

Abstract:

Whether compounds are intended as drugs, cosmetics, agrochemicals or for other industrial application, it is essential to understand their potential to cause toxic effects. This can guide the prioritisation of compounds for further research or consideration of the most appropriate downstream experiments to confirm their safety. The ability to predict toxicities based on chemical structure alone would allow these factors to be considered prior to synthesis, allowing the safest options to be pursued and saving time and resources wasted on synthesis and testing of unsuitable compounds.

This is a preprint of an article that will appear in J. Chem. Inf. Model. http://dx.doi.org/10.1021/ci2003208.

This describes the concepts and algorithms underlying StarDrop's Nova module.

Abstract

In this article we describe a computational method that automatically generates chemically relevant compound ideas from an initial molecule, closely integrated with in silico models and a probabilistic scoring algorithm to highlight the compound ideas most likely to satisfy a user-defined profile of required properties. The new compound ideas are generated using medicinal chemistry ‘transformation rules’ taken from examples in the literature. We demonstrate that the set of 206 transformations employed is generally applicable, produces a wide range of new compounds and is representative of the types of modifications previously made to move from lead-like to drug-like compounds. Furthermore, we show that more than 94% of the compounds generated by transformation of typical drug-like molecules are acceptable to experienced medicinal chemists. Finally, we illustrate an application of our approach to the lead that ultimately led to the discovery of Duloxetine, a marketed serotonin reuptake inhibitor.

We've just submitted this review article, "Multi-Parameter Optimization: Identifying high quality compounds with a balance of properties" to a special issue of Current Pharmaceutical Design. In it, we survey the range of methods used for MPO in drug discovery, compare their strengths and weaknesses and present some example applications.

Matt gave this presentation at the ACS Spring meeting 2011 in Anaheim.

Abstract

Computational tools can guide the selection of high quality compounds, with a good balance across multiple properties, and guide strategies to design improved compounds. But, can software propose ideas for better compounds? We will demonstrate an approach that generates compound ideas and identifies those that are most likely to achieve a drug discovery project’s objectives. The compound ideas should be synthetically feasible; to achieve this, new structures are generated from an initial compound using medicinal chemistry ‘rules’, using a method similar to [1]. These are then scored against a profile of property criteria using a probabilistic scoring method [2] and visualized in ‘chemical space’ to allow many ideas to be rapidly explored and prioritized for detailed consideration.

[1] Stewart et al. Bioorg. Med. Chem (2006), 14 p. 7011

[2] Segall et al. Chem. & Biodiv (2009), 6 p. 2144

Matt gave this presentation at the ACS Spring meeting 2011 in Anaheim.

Abstract

When we build a predictive model of a drug property we rigorously assess its predictive accuracy, but we are rarely able to address the most important question, “How useful will the model be in making a decision in a practical context?” To answer this requires an understanding of the prior probability distribution and hence prevalence of negative outcomes due to the property. We will illustrate the importance of the prior to assess the utility of a model to select or eliminate compounds for further investigation. A better understanding of the prior probabilities of adverse events due to key factors will improve our ability to make good decisions in drug discovery, finding higher quality molecules more efficiently. As the data necessary to estimate these priors does not include proprietary compound structures, this presents an opportunity for collaboration to improve the basis for good decision-making for all.

This whitepaper describing Nova's technology has not yet been published.

Abstract

Using in silico predictive models and multi-parameter optimisation techniques allows large numbers of compounds to be quickly assessed with respect to a profile of properties required for a successful compound in a drug discovery project. With these predictive methods, it becomes possible to consider a large number of ideas for potential compounds that can be easily created and entered into a computer by an individual. In this article we describe a method that automatically generates chemically relevant compound ideas from an initial molecule, based on medicinal chemistry ‘transformation rules’ taken from examples in the literature. These are then prioritised using in silico models and a probabilistic scoring algorithm to identify the compound ideas most likely to satisfy a user-defined profile of required properties. Embedded in an intuitive, visual user interface, this approach provides a powerful means to explore potential chemistry to identify high quality leads or to improve properties in lead optimisation. We demonstrate that the set of 206 transformations employed is generally applicable, produces a wide range of new compounds and is representative of the types of modifications previously made to move from lead-like to drug-like compounds. Furthermore, we show that more than 94% of the compounds generated by transformation of typical drug-like molecules are acceptable to experienced medicinal chemists. Finally, we illustrate an application of our approach to the lead that ultimately led to the discovery of Duloxetine, a marketed serotonin reuptake inhibitor. Our analysis results in the identification of a diverse range of high scoring compounds, including Duloxetine itself.

This article was published in J. Comp. Aided Mol. Design (DOI 10.1007/s10822-010-9388-7) and discusses a critical issue that the community needs to address address in order to use the predictive models that we build to the greatest effect.

Abstract

When we build a predictive model of a drug property we rigorously assess its predictive accuracy, but we are rarely able to address the most important question, “How useful will the model be in making a decision in a practical context?” To answer this requires an understanding of the prior probability distribution (“the prior”) and hence prevalence of negative outcomes due to the property being assessed. In this perspective, we illustrate the importance of the prior to assess the utility of a model in different contexts: to select or eliminate compounds, to prioritise compounds for further investigation using more expensive screens, or to combine models for different properties to select compounds with a balance of properties. In all three contexts, a better understanding of the prior probabilities of adverse events due to key factors will improve our ability to make good decisions in drug discovery, finding higher quality molecules more efficiently.

Matt presented this poster at ISSX in September 2010.

Abstract:

A number of methods have been developed for the prediction of regioselectivity of metabolism by the major drug metabolising isoforms of Cytochrome P450 [1,2,3]. However, while valuable, predicting the relative proportion of metabolite formation at different sites on a molecule is only a partial solution to designing more stable molecules. Valuable additional information comes from predicting a measure of the vulnerability of each site to metabolism. Such a measurement is the site lability, as calculated by StarDrop. This important factor in determining the overall rate of metabolism, when combined with other descriptors relating to substrate affinity, can provide good predictive models of in vitro metabolic rate which can, in turn, guide design of compounds with improved stability.

Dr Terry Stouch gave this presentation as part of the first StarDrop User Group Meeting and Workshop at the ACS Fall meeting 2010 in Boston.

Matt presented this poster at MipTec in September 2010.

Abstract:

People are notoriously poor at making good decisions based on complex, uncertain data when there is a lot at stake. In drug discovery, poor decisions can mean wasting effort in synthesizing and testing compounds that fail or throwing out perfectly good compounds in error, reducing the opportunities to find new, valuable therapies. However, making good decisions in this context is challenging for several reasons: the need to balance multiple, often conflicting criteria for a successful drug; the abundance of data available on many properties; and the uncertainty in the relevance and accuracy of the available data, particularly in early discovery.

In this presentation Ed Champness demonstrates some of StarDrop's features for identifying and selecting compounds, using chemical space representations to illustrate two case studies.

Abstract:

The term chemical space is generally understood to describe the universe of possible chemistries that may exist. With this number being almost incomprehensible, the term is more often used in a specific company or project context where the bounds are more manageable both theoretically and computationally. In this paper we present a method for constructing and visualising chemical space for use in a project setting. When exploring this space, we need to consider how to balance quality and diversity, or alternatively, ‘exploration’ and ‘exploitation’, and how to avoid introducing bias into our selections. It is often important not to focus too quickly, but instead to gather data on potential backup series in order to mitigate risk and understand the SAR. With this in mind, basing our selections purely upon a single property, such as the potency of the compounds, can lead us towards very different choices that may ultimately result in more complex challenges as we attempt to design leads which have a good overall balance of properties. We present examples and case studies to illustrate these concepts and results from both Hit-to-Lead and Lead Optimisation.

Ed gave this presentation at MipTec, September 2010, in Basel.

Dr Mikel Moyer gave this presentation as part of the first StarDrop User Group Meeting and Workshop at the ACS Fall meeting 2010 in Boston.

Dr Terry Stouch, Consulting in Drug Discovery and Design Practice, Technologies, Process at Princeton, NJ and Duquesne University gave this presentation on "In silico ADME/Tox: Why models fail: Why models work", in June 2010.

Abstract:

By way of example, we discuss the apparent "failure" of in silico ADME/Tox models and attempt to understand the causes. Often,the interpretation of the success of models lies in their use and the expectations of the user. Other times, models are, in fact, of little value. Disappointing results can be linked to the key aspects of the model and modeling procedure, many of these related to the original data and its interpretation. We make recommendations to providers of models regarding the development, description, and use of models as well as the data and information

Terry gave this presentation at the 32nd National Medicinal Chemistry Symposium, June 6-9th, 2010, Minneapolis, MN, USA.

A copy of Terry's slides is available as a PDF file.

This paper was published by Andrew Chadwick and Matthew Segall in Drug Discovery Today, 2010, 15 (13/14), pp. 561-569.

Abstract

Better individual and team decision-making should enhance R&D performance.  Reproducible biases affecting human decision making, known as cognitive biases, are well understood by psychologists. These threaten objectivity and balance and so are credible causes for continuing unpleasant surprises in Development, and high operating costs.  For four of the most common and insidious cognitive biases, we consider the risks to R&D decision-making and contrast current practice with use of evidence-based medicine by healthcare practitioners.  Feedback on problem solving performance in simulated environments could be one of the simplest ways to help teams improve their selection of compounds and effective screening sequences. Computational tools that encourage objective consideration of all of the available information may also contribute.

The published article can be accessed at http://dx.doi.org/10.1016/j.drudis.2010.05.007 or you can download a preprint free of charge.

Dr Andrew Chadwick, Consultant (Life Sciences and Healthcare) at Tessella gave this presentation on "Rational Approach to Risk Reduction: What can discovery screening planners learn from volcanos and dust detection?", on Wednesday 6th June 2010, covering the following topics:

• The purpose of screening and the principles of value-adding screening plans
• The importance of finding the right, tailored screening plan for each project
• Risk perception and the pitfalls of cognitive biases for decision-making
• Metrics that should guide the choice, sequence, combinations and cut-offs for tests
• Ways of balancing the important factors (downstream consequences of error, cost and time for screening, and predictive performance)
• The need to overcome the challenge of uncertain inputs
• What is the right balance between exploitation and exploration of product options and technology performance?
• Effective approaches to supporting learning and continuous improvement

Andrew gave this presentation at the 11th Annual Drug Discovery Leaders Summit, June 9-10th, 2010, Montreux, Switzerland.

This paper was published by Olga Obrezanova and Matthew D. Segall, Journal of Chemical Information and Modeling, 2010, 50 (6), pp 1053–1061.

Abstract

In this article, we extend the application of the Gaussian processes technique to classification quantitative structure−activity relationship modeling problems. We explore two approaches, an intrinsic Gaussian processes classification technique and a probit treatment of the Gaussian processes regression method. Here, we describe the basic concepts of the methods and apply these techniques to building category models of absorption, distribution, metabolism, excretion, toxicity and target activity data. We also compare the performance of Gaussian processes for classification to other known computational methods, namely decision trees, random forest, support vector machines, and probit partial least squares. The results indicate that, while no method consistently generates the best model, the Gaussian processes classifier often produces more predictive models than those of the random forest or support vector machines and was rarely significantly outperformed.

You can donwload it via the ACS "Articles on Request" e-print service using the following link:

http://pubs.acs.org/articlesonrequest/AOR-e9ivq4kdydbtNGCKt9Nw

Please note: To access the Articles on Request link, please log in to the Publications website using your ACS ID. If you do not have an ACS ID, you will need to Register for one for free by clicking on 'Register' near the top right corner of the website.

In this presentation Ed Champness considers the decision-making challenges faced by drug discovery scientists and presents some visual analyses that can be used to help answer some of the common questions that are asked.

Ed gave this presentation at the ACS Spring meeting 2010 in San Francisco.

In this article Olga describes how we extend the application of Gaussian Processes technique to classification problems. We explore two approaches, an intrinsic Gaussian Processes classification technique and a probit treatment of the Gaussian Processes regression method. Here we describe the basic concepts of the methods and apply these techniques to building category models of blood-brain barrier penetration and hERG inhibition. We also compare performance of Gaussian Processes for classification to other known computational  methods, namely decision trees, bagging and probit PLS.

In this presentation Wolf Ihlenfeldt describes some of Xemistry's core competencies.

Xemistry is one of Optibrium's partners, providing the CACTVS toolkit which is used to manage the underlying computational chemistry behind StarDrop.

Young Shin and his colleagues at Genentech presented this poster at the ISSX North American Regional meeting in Baltimore, MD in October 2009.

COMPARISON OF METASITE AND STARDROP PREDICTION OF CYP3A4, CYP2C9 AND CYP2D6 V. Sashi Gopaul, Young Shin, Hoa Le, Matthew Baumgardner, Cornelis Hop and Cyrus Khojasteh Drug Metabolism & Pharmacokinetics, Genentec, Inc, South San Francisco, CA, USA, 94080

Metabolite identification studies play an important role in determining the sites of metabolic liability of new chemical entities (NCEs) in drug discovery. However, generating these complex and detailed studies in a highthroughput environment is often a challenge. Therefore, the use of in silico tools that can predict the sites of metabolism of an NCE could enhance the drug design process. In this study we compare the utility of MetaSite and Stardrop, two predictive softwares available for this purpose...

This article was published in Chemistry & Biodiversity, Volume 6, Issue 11, Pages 2144 - 2151

Abstract

ADMET models, whether in silico or in vitro, are commonly used to ‘profile’ molecules, to identify potential liabilities or filter out molecules expected to have undesirable properties. While useful, this is the most basic application of such models. Here we will demonstrate how models may be used to go ‘beyond profiling’ to guide key decisions in drug discovery. For example, selection of chemical series to focus resources with confidence or design of improved molecules targeting structural modifications to improve key properties....

Dan Ortwine gave this presentation as part of a workshop at the ISSX North American meeting in Baltimore, MD, USA, 2009.

The workshop, titled Effectively Using In Vitro Data, In Silico Models and Data Mining at Early Stage Drug Development had the following introduction:

At the early stages of drug discovery there are large numbers of compounds synthesized to examine the chemical landscape needed for discovering the lead compounds. Chemical diversity plays a key role at this stage to increase the probability of success when narrowing down to clinical candidates. In this workshop, we will discuss a number of currently available in vitro assays and in silico tools and how they are complementary. We hope by the end of the workshop to provide a platform for decision making purposes in early drug discovery.

Defining a property profile is subjective and often leads to lengthy, interdisciplinary discussions about the criteria and their relevance. For example, is it worth sacrificing some potency to gain additional metabolic stability or solubility? However, a question that is rarely asked is, “What impact would that trade-off have on the final outcome?”, particularly given the underlying uncertainty.

This poster was displayed at the ISSX North American meeting in Baltimore, USA, 2009.

Matthew Segall (while still in the ADMET Division, BioFocus DPI), explored the balance between luck and judgement in drug discovery. As Matt put it "The vision of an in silico design process for drug molecules is certainly attractive, so
why has this goal yet to be realised, despite an enormous effort over the past 10 years?"

In this presentation Olga Obrezanova describes  an automated process for building QSAR models (now available as part of StarDrop as the Auto-Modeller!). Olga goes on to demonstrate the effectiveness of the process by carrying out comparisons of this technique with traditional "hand-on" modelling approaches for blood-brain barrier penetration and aqueous solubility.

This presentation was given at the Zing Computational Chemistry Conference in March 2009.

The current paradigm of  drug discovery utilising chemical library synthesis coupled with high throughput screening technologies often gives rise to a situation whereby drug discovery programmes are compound rich although poor in ADME properties. As such, the ADME properties of compounds require optimisation, through the phases of lead optimisation, prior to progression for clinical development increasing the cost and duration of the process. A prime driver of drug discovery is therefore the early identification of compounds from diverse chemical spaces with optimal ADME properties.

This poster was displayed at the ISSX international meeting in Japan, 2007.

This article was published in QSAR & Combinatorial Science, Volume 25, Issue 12, Pages 1172 - 1180 (DOI
10.1002/qsar.200610093)

Abstract
In this article, we review recent developments in the prediction of Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties by Quantitative Structure – Activity Relationships (QSAR). We consider advances in statistical modelling techniques, molecular descriptors and the sets of data used for model building and changes in the way in which predictive ADMET models are being applied in drug discovery. We also discuss the current challenges that remain to be addressed. While there has been progress in the adoption of non-linear modelling techniques such as Support Vector Machines (SVM) and Bayesian Neural Networks (BNNs), the full advantages of these "machine learning" techniques cannot be realised without further developments in molecular descriptors and availability of large, high-quality datasets. The largest pharmaceutical companies have developed large in-house databases containing consistently measured compound properties. However, these data are not yet available in the public domain and many models are still based on small "historical" datasets taken from the literature. Probably, the largest remaining challenge is the full integration of predictive ADMET modelling in the drug discovery process. Until in silico models are applied to make effective decisions in a multi-parameter optimisation process, the full value they could bring will not be realised.

This is a preprint of the article published in J Comput Aided Mol Des. 2008 Jun-Jul;22(6-7):431-40. Epub 2008 Feb 14.

Abstract

In this article, we present an automatic model generation process for building QSAR models combined with Gaussian Processes, a powerful machine learning modeling method. We describe the stages of the process that ensure  models are built and validated within a rigorous framework: descriptor calculation, splitting data into training, validation and test sets, descriptor filtering, application of modeling techniques and selection of the best model. We apply this automatic process to data sets of blood-brain barrier penetration and aqueous solubility data sets and compare the resulting automatically generated models with ‘manually’ built models using external test sets. The results demonstrate the effectiveness of the automatic model generation process for two types of data sets commonly encountered in building ADME QSAR models, a small set of in vivo data and a large set of physico-chemical data.

In this presentation, Alan Beresford discusses drug discovery and considers how ADME models fit into the process as a step necessary for helping to manage the numbers game. Alan describes how, by using appropriate interpretation of ADME model results, it is possible to credibly include them within a traditional screening cascade.

This presentation was given at the MedChem USA Conference in 2007.

The rapid design-test-redesign cycles of modern drug discovery and the demand for fast model (re)building whenever data becomes available have given rise to a trend to develop computational algorithms for automatic model generation. Automatic modelling processes allow computational scientists to explore large numbers of modelling approaches very efficiently and make QSAR/QSPR model building accessible to non-experts.

This poster was displayed at MedChem ADMET Eurpoe, 2008.

In silico predictive models are now widely used to predict a range of molecular properties and help prioritise molecule for synthesis. However, a common criticism often levelled at predictive models is that they offer few clues regarding why a molecule is predicted to have a certain property. By definition, models encode relationships between molecular structure and properties, but interpreting and visualising this information to design better molecules has been almost impossible. This is particularly true of models built with modern ‘machine learning’ techniques such as artificial neural networks (ANN), Gaussian processes (GP) or support-vector machines (SVM). The models that these techniques create have commonly been described as ‘black box.’

This poster was presented at the MedChem Europe meeting, 2007.

In this presentation, Matt Segall talks about the differences between "design" and "discovery" and considers two different analogies for the drug design process - the development of the Boeing 777 and card counting in blackjack. Using the latter, Matt discusses how we can appropriately use uncertain information to guide decisions amd how we can interpret in silico data to guide compound design. Finally, Matt gives an illustrative example of putting this theory into practise in a case study during which the aim was to focus resources in a hit-to-lead/lead optimisation study.

This presentation was given at the SMI In Silico ADMET conference in 2007.

In this presentation Olga Obrezanova talks about Gaussian Processes - a powerful computational method  for QSAR modelling. Olga starts by describing the main ideas of this technique.

We are mostly interested in the application of this technique to predictive modelling of ADME properties. The importance of optimising ADME properties of potential drug molecules is now widely recognised. Considering the ADME properties early in the drug discovery process can reduce the costs of the drug development and decrease the attrition rate of drug candidates. We have developed new techniques for finding parameters of the Gaussian Processes method which I will present. I will also show examples of application of these techniques to ADME and QSAR datasets and compare Gaussian Processes methods with other known techniques. The demand of modern drug discovery for fast model (re)building whenever new data becomes available gave rise to a trend to develop computational algorithms for automatic model generation. I will demonstrate how we use Gaussian Processes in an automatic modelling process. (The purpose of such algorithms is to save scientists' time, explore more modelling possibilities and make the process of QSAR model building accessible to non-experts.)

This presentation was given at the American Chemical Society conference in Boston, 2007.