For the first time we are holding a predictive ADME and Toxicology workshop in Oxford this summer. It should be a valuable and rewarding experience that adds to our summer workshop activities there. I believe we have a really good group of facilitators gathered that should make for an excellent working week. We will take a working, problem-solving approach to case study datasets throughout the week.
Here is the program as a pdf download:
Download ECheminfoADMETProgramOxford09
More detail on the program abstracts and schedule in the continuation below.
Barry
Monday July 27
08.30 Registration Open
09.00 Overview of Workshop Activities, Barry Hardy (Douglas Connect)
09.15 Group Introductions
10.00 Library Design incorporating ADME Prediction and Metabolic Properties, Ismael Zamora (Pompeu Fabra University and Lead Molecular Design)
Absorption, Distribution, Metabolism and Excretion (ADME) properties are considered in the designing of new compounds during the drug discovery process. It is not enough that a compound is active in the target protein of interest, the potential new drug also has to reach the site of effect and stay there enough time to elicit the pharmacological activity. Therefore the prediction of these properties is used in the discovery of new potential drugs to select and design more adequate compounds in a faster way. The aim of this workshop is to show and practice several methods and tools for the prediction of ADME properties.
Absorption and distribution properties including permeability, solubility, un-specific protein binding and volume of distribution will be studied using VolSurf+ software. Metabolism properties will be analyzed using the MetaSite 3.0 software that considers the cytochrome-mediated metabolism of xenobiotics.
The model system used in this study will be the nonsteroidal anti-inflammatory drug celecoxib, a COX-2 selective inhibitor and known CYP2C9 substrate.
In this workshop the participants will learn which are the most relevant ADME properties and how to design compounds to overcome the potential problems that may slow down the lead identification and optimization processes.
13.00 Lunch
14.00 Using Database look-ups, Read Across, and Predictive Models to assess the Toxicity of Compounds, Glenn Myatt (Leadscope)
In silico toxicology software programs are becoming increasingly useful throughout all stages of drug discovery, including lead identification, lead optimization, and preclinical. Tools such as database look-ups, read across, and predictive models can be used to assess the toxicity of compounds. The use of these tools at the various stages of the discovery process will be explored in this workshop. The Leadscope software and databases will be used to provide hands-on experience of these different approaches. The software provides a range of structure and toxicity searching options over a variety of toxicity databases, including data from the US FDA. The software will be used to find information on compounds, to perform read across, and make predictions. Issues concerning data mining toxicity information will be further examined including finding structural alerts, building predictive models, and the use of modeling where the training data can or cannot be examined.
17.00 Group Work and Discussion on Workshop Case Study Problems
18.00 End of Workday
19.00 Refreshments and Food (St Edmund Hall)
Tuesday July 28
8.30 Knowledge-based Reasoning for the Prediction of Toxicity and Metabolism, Philip Judson (Lhasa Ltd.)
This workshop will look at non-statistical approaches to the prediction of toxicity and metabolism. Programs from Lhasa Limited for predicting toxicity (Derek for Windows) and mammalian metabolism (Meteor) will be used for illustration but the main aim of the workshop will be to explore the scientific case for reasoning-based prediction and its implications for computer methods.
We put trust in reasoning-based prediction in everyday life without even noticing that we are doing it – for example, to prepare for heavy traffic on the way to work if there is to be a test match at a cricket ground en route. Is our innate trust in reasoning methods justified? How objective is scientific thinking? Might there be a place for subjectivity? How do reasoning-based methods compare with statistical ones in terms of scientific validity and reliability? How can you assess the reliability of a qualitative prediction and how satisfactory are the accepted methods for assessing the reliability of statistical predictions?
The workshop will look at how human experts use reasoning to make predictions about chemical toxicity and xenobiotic metabolism, and how some of their thinking has been incorporated into knowledge-based computer programs. Reference will be made to knowledge-based programs for predicting toxicity and metabolism other than those developed by Lhasa Limited, and to some for making predictions in related fields such as ecotoxicity and environmental biodegradation. The extent to which statistical and reasoning-based methods are complementary or contradictory will be discussed.
Workshop participants should gain an understanding of how human knowledge is incorporated into knowledge-based systems and how to judge and compare predictions from reasoning-based and statistical models.
11.30 Group Work on Workshop Case Study Problems
12.30 Lunch
13.30 Lazy-Structure-Activity-Relationships (lazar) for the in-silico Prediction of Chemical Carcinogenicity, Christoph Helma (in silico toxicology)
In this workshop the group will apply the lazar (Lazy Structure Activity Relationships) system for the prediction of biological activities and its application for the prediction of carcinogenicity, an endpoint that is very hard to predict with existing (Q)SAR techniques.
lazar uses a modified k-nearest-neighbor algorithm, that is capable of detecting activity specific chemical similarities, to derive predictions for untested structures from a database with experimental toxicities. lazar relies on relatively few model assumptions and provides the rationales for predictions in an understandable and traceable manner. The system is capable of discriminating reliably between trustworthy and untrustworthy predictions (e.g. for structures that fall beyond the scope of the training set) by assigning a confidence index to each prediction.
The group will carry out cross-validation experiments with various carcinogenicity and mutagenicity endpoints to determine the predictive accuracies for structures within the applicability domain of the training data. The group will determine where lazar can reliably identify cases where the information in the database is insufficient and/or contradictory to derive valid predictions.
16.30 Group Work on Workshop Case Study Problems
17.30 End of Workday
18.00 Punting
Wednesday July 29
8.30 Modelling the “Toxicity Pathway” from Chemistry to Effect, Mark Cronin (School of Pharmacy and Chemistry, Liverpool John Moores University)
One of the simplest and powerful approaches to predict toxicity is to group chemicals together into categories. If the group, or category, can be populated with reliable and relevant existing data, read-across can be performed to make predictions of toxicity. Whilst the concept is simple, and increasingly applied, the key is to form reliable categories. This workshop will illustrate the formation of categories from structural analogues, mechanisms of toxic action, as well as chemicals with similar modes. This will be illustrated through the freely available OECD (Q)SAR Application Toolbox.
In this workshop the group will:
• Be given an introduction to the background and philosophy of the OECD (Q)SAR Toolbox
• Obtain hands-on experience of the Toolbox to illustrate:
o Basic functionality
o Chemical structure input
o Data retrieval facilities
o Category formation
o Read-across
• Carry out category formation, read-across and analogue approaches to predicting toxicity
• Model the “toxicity pathway” from chemistry to effect
11.00 Group Work on Workshop Case Study Problems
12.00 Lunch
13.00 QSAR Validation Criteria and Concepts, Judith Madden (School of Pharmacy and Chemistry, Liverpool John Moores University)
For confidence to be assigned to a predictive approach in toxicology, there must be assessment of the robustness and accuracy of a model. The OECD Principles for the Validation of (Q)SARs and associated guidance have been developed precisely for this purpose. These principles provide a framework to verify, characterise, evaluate and possibly validate (Q)SARs and predictive toxicology approaches. In this workshop an introduction will be given to the concepts of QSAR assessment. The group will perform a variety of exercises based on the OECD validation criteria. Concepts will be presented to characterise various in silico toxicology approaches. The relevance of these approaches, originally designed for REACH endpoints, to a broader context such as pharmaceuticals, will be illustrated.
15.30 Group Work on Workshop Case Study Problems
16.30 Application of Structure-Activity Relationships in REACH-compliant Chemical Hazard Assessment, Arianna Bassan (S-IN)
This training session focuses on the possible use of non-testing methods in the regulatory assessment of chemicals especially in relation to the REACH legislation, which came into force in Europe on 1st June 2007. This new EU regulatory framework aims at improving the protection of human health and environment through the better and earlier identification of the properties of chemical substances.
In the regulatory framework there is a growing need for in silico methods that can be used to gain information about the environmental fate and ecological and health effects of chemicals. The different techniques that are used to derive non-testing information include (quantitative) structure-activity relationship models, expert systems, and read-across/category approaches. National and international agencies have a number of reasons to encourage the use of in silico methods. First of all, computational methods are faster and cheaper compared to empirical testing methods, and their use results in considerable savings of time and money during the assessment of chemical hazard. To limit the cost and the number of animals used for testing, REACH explicitly encourages the use of computer-aided methods such as (Q)SAR methods and category/read-across approaches for filling in the enormous knowledge gap of chemical information. In order to be used in place of experimental data, REACH requires that the in silico methods meet certain conditions. For example, in the case of (Q)SARs, these requirements include: 1) the model has to be valid; 2) the substance has to fall within the applicability domain; 3) the prediction has to be adequate for the regulatory purpose; 4) the applied method has to be provided with adequate and reliable documentation.
The use of in silico methods, and in particular, (Q)SARs, as valuable components of the regulatory assessment strategy is hampered by two major factors. First, model estimations can be properly interpreted only by specialists with specific expertise in the field of computational toxicology, and this factor certainly limits the widespread use of in silico approaches for regulatory purposes. Second, for a method to be accepted in a regulatory framework, its scientific validity has to be established in accordance with internationally agreed validation principles. At present the scientific validity of many models is not documented, which then limits their regulatory acceptance.
In this training session, the following topics will be presented and discussed:
• Introduction on the REACH regulation (and in particular Annex XI) and the OECD Principles for (Q)SAR Validation.
• Regulatory use of QSARs in the REACH framework
• Reporting Formats (e.g. QSAR Model Reporting Format, QMRF) for providing adequate documentation about the models.
• Structured workflow that assists users all the way through the generation of reliable non-testing data and by aiding the following processes:
o Retrieving existing physicochemical properties and (eco)toxicological information for a given chemical.
o Selecting relevant in silico approaches for predicting individual toxic endpoints.
o Generating endpoint predictions.
o Providing information on the reliability of the estimates.
o Exploiting the capability of various in silico methodologies.
o Integrating results.
o Compiling robust summaries that document in a transparent way the use of the methods.
• Review of computational tools for applying (Q)SARs.
• Hands-on session (e.g. use of selected tools such as Toxmatch).
19.00 Group Work on Workshop Case Study Problems
20.00 End of Workday
Thursday July 30
8.30 Modelling Skin Penetration, Irritation, Sensitization and Penetration Enhancement, Tony Hopfinger (University of New Mexico)
The modeling of an ADMET endpoint is highly dependent upon the complexity of the molecular mechanism involved. In cases where the molecular mechanism is complex, and/or pharmacological understanding is quite limited, an empirical informatics approach to develop predictive models is the preferred methodology to apply. Unfortunately, for ADMET studies, few high-level 3D and 4D level descriptors are available to model complex mechanisms of action as compared to drug potency QSAR investigations. We have developed a set of universal descriptors, called 4D-fingerprints, 4D-FP, which capture the three-dimensional size, shape, chemical composition, reactive state and molecular flexibility of a molecule for informatics type ADMET modeling. For ADMET endpoints where cellular membrane permeation and diffusion are involved, a pseudo structure-based design approach called membrane-interaction (MI-) QSAR analysis can be applied. Here descriptors derived from the simulation of an organic molecule passing through a phospholipid membrane assembly are used with intramolecular descriptors derived from the organic molecule to build MI-QSAR models. With our most recently developed MI-QSAR descriptor, the difference in the integrated cylindrical distribution functions over a phospholipid monolayer model, in and out of the presence of a monolayer penetrator, greatly reduces the size and complexity of the MI-QSAR models as compared to corresponding classic intramolecular QSAR models. In this workshop participants will study "Skin" ADMET endpoint modeling, using both 4D-FP and/or MI-QSAR descriptors in tandem with the more 'traditional' 1D and 2D intramolecular descriptors. The specific skin ADMET endpoint 4D-FP and MI-QSAR case study applications will be:
a) skin penetration - which is desired for transdermal delivery, but may involve unwanted irritation and sensitization
b) skin irritation and sensitization - requires study of both transport and chemical reactivity
c) skin penetration enhancement - a desired property for cosmetics and drugs, but may involve toxicity issues
d) therapeutic index modeling - optimizing penetration and minimizing irritation, sensitization and/or toxicity as a function of chemical structure
11.30 Group Work on Workshop Case Study Problems
12.30 Lunch
13.30 ADME QSAR Modelling to Guide Drug Design, Olga Obrezanova (BioFocus DPI)
The importance of optimizing ADME properties of potential drug molecules is now widely recognized. In silico predictive modelling offers a possibility to consider the likely ADME properties of a molecule before it is even synthesized. This workshop will focus on building predictive models and using them to guide drug design. The participants will build a QSAR/QSPR model of a property using an algorithm for automatic model generation based on Gaussian Processes, a powerful ‘machine learning’ technique. The predictions from this model will be used alongside other compound data, including in vitro measurements and in silico predictions from other ADME QSAR models, to identify compounds with a balance of appropriate ADME and potency. Using a ‘probabilistic’ scoring algorithm, the participants will be able to prioritise and select compounds most likely to meet the project criteria. The participants will see how a predictive model coupled with a visualisation tool, which provides a link between compound structure and predicted property values, can help to guide the redesign of compounds to overcome liabilities.
The workshop will be based on the StarDrop™ software platform which helps to guide decisions for compound optimization, design and prioritisation. The participants will have access to different functionalities within StarDrop: a suite of predictive ADME QSAR models, a model building tool (Auto-Modeler), a visualization tool (Glowing Molecule), and the unique probabilistic scoring algorithm which is able to rapidly integrate all the compound data, predicted and experimental, to prioritize compounds with the best balance of properties.
16.30 Group Work on Workshop Case Study Problems
18.00 End of Workday
Friday July 31
9.00 Evaluation of Strengths and Weaknesses of QSAR-based Predictive ADMET Workflows, David Leahy (University of Newcastle)
Whereas good quality ADMET data was once difficult to obtain it is now routinely generated alongside drug discovery programs for a significant proportion of compounds of interest and large companies have accumulated datasets for up to 10,000 and more compounds across multiple chemical series and multiple assays. This creates an opportunity for us to explore both generic and local QSAR modelling methods that can be routinely updated as new data is added. The workshop will demonstrate best-practice QSAR modelling approaches using Inkspot Science's online integration platform to simplify access to the best community chemoinformatics and statistical tools. We will also evaluate the strengths and weaknesses of alternative QSAR modelling methods using the competitive workflow methods of the "Discovery Bus", a novel system for automating multiple informatics methods.
12.00 Lunch
13.00 Group Work on Workshop Case Study Problems
16.00 Group Discussion of Workshop Case Study Results
17.00 End of Workshop
Comments