- February 2001 Pharmaceutical Research & Development is going through a paradigm shift as the IT revolution provides us with fast, effective and more sophisticated ways of generating and processing R&D data from discovery to clinical research. The business has irreversibly changed and the companies who want to meet the new challenges will have to embrace the new technologies in order to increase productivity, cut costs and reduce time to market. InfoTechPharma is THE Information Technology event for the Pharmaceutical Industry. Over the last four years InfoTechPharma has provided hundreds of participants the opportunity to hear cutting edge advice on how to harness the power of IT to provide real competitive advantage across all disciplines. Please find three Executive Summaries from papers presented at IBC Life Sciences, InfoTechPharma Congress. Please click on one of the below options to view the Executive Summary written from that presentation. Development of pharmacogenomic models to predict drug response e-Readiness of R&D: Addressing Knowledge and Informaiton Management Competence QSARs and Expert Systems to Predict Toxicity Discovering and Managing Knowledge in HTS datasets "Genetics and Computing: New Technologies To Improve Drug Discovery and Research" For further information and details of the comprehensive documentation available from this event, please visit: www.Infotechpharma.com Details of InfoTechPharma 2002 will be available soon, please visit: www.Infotechpharma.com 1. Development of pharmacogenomic models to predict drug response [top] Gail I.R. Adam Ph.D., Head of Molecular Science, Gemini Genomics, Sweden. Individuals show variation in responses to particular drug treatments. That such variations could be attributed to genetic factors was first demonstrated several decades ago, but it is only now that we are beginning to understand how and why. Mean average approach to treatment efficacy: Population-wide drug responses typically appear as Gaussian distribution curves, with some patients responding well and others poorly or adversely to particular drugs or drug classes. The pharmacogenomic approach allows stratification of patient populations to achieve better efficacy of treatment, while also providing opportunities for rationalization of drug trials and identification of new markets. Genomic sequence and polymorphism information: The assessment of the role of genetic factors in drug response can be carried out at the gene, transcript or protein level. Public and private genomic sequencing and in particular, SNP determination efforts, have greatly increased the current potential of pharmacogenomics and reduced time and costs involved in the hunt for both functional genetic variants and surrogate markers. Investigations at the genome level are predominantly used today, but future combinations of pharmacogenomics and proteomics approaches may yield the most powerful predictive tools for drug response evaluation. Good clinical data is essential: In order to determine the contribution of particular genetic variants to drug responses, it is of paramount importance that clinical measurements are accurate. This can be particularly difficult for complex quantitative trait disorders where environmental, as well as genetic factors have to be considered. Predictive models and principal component analysis: Training sets, test sets and independent prediction sets of data require examination. In addition to response prediction models, replication and validation studies need to be carried out. Spurious correlations between genetic and phenotypic data should be expected and population cohorts drawn from different geographical locations where possible, in order to minimize genetic background type-1-error risks. 2. e-Readiness of R&D: Addressing Knowledge and Informaiton Management Competence [top] Ms Elisabeth Goodman, Glaxosmithkline, UK Data, information and knowledge are key inputs and outputs in the Pharmaceutical R&D process. Information Technology is increasingly geared to enabling end-users (R&D scientists, clinicians, managers) to generate, access and manipulate these R&D assets. As a result, many are experiencing difficulties in knowing where best to go for quality information, and how to avoid information overload. Information professionals will continue to provide expert information consultation and analysis, as well as coaching and support in the use of web and other desk-top information sources. But, as R&D practitioners, you will increasingly need to develop your own competencies to optimise your use of these key R&D knowledge and information assets. This paper describes the competencies needed and, through a case study, what could be done to establish them in Pharmaceutical R&D. 3. QSARs and Expert Systems to Predict Toxicity [top] Mark Cronin, School of Pharmacy and Chemistry, Liverpool John Moores University The elucidation of toxicity is perceived as one of the most fundamental bottlenecks in the drug discovery process. As such, early identification of toxicity is crucial. A potential solution to the problem of identifying toxicity, which negates the requirement for costly and time-consuming animal testing, is the application of computer-aided toxicity prediction. There are two approaches to computer-aided toxicity prediction, namely quantitative structure-activity relationships (QSARs) and expert systems. These have the potential to predict a range of toxicities from a knowledge of physico-chemical structure alone. QSARs are available to predict a wide range of toxic endpoints. Such mathematical relationships should be based upon a mechanistic understanding of the toxic process, namely that transport to the putative site of toxic action, and interaction at that site of action should be modelled. There is an exhaustive literature available on toxicological QSARs and many of the results have been formalised into databases. Expert systems for toxicity prediction enable a user to make rational predictions about the toxicity of chemicals. Broadly speaking expert systems can be split into two categories. Some, such as TOPKATT and related systems are effectively automated QSARs. Others are based upon the application of a knowledge base in the form of 'rules'. Expert systems for toxicity prediction such as DEREKT, StART, and HazardExpertT are based upon rules derived by human experts. Other systems such as CASET use algorithms to derive these rules automatically. A variety of expert systems exist, and each with its own strengths and weaknesses. Computer-aided prediction of toxicity can be applied at a number of stages in the drug development process. Many systems are available to provide a general screen for toxicity that could be utilised concomitantly in the early stages of drug discovery e.g. to assess toxicity for large numbers of compounds produced by combinatorial chemistry. Other systems are applicable for the rational development of non-toxic lead compounds. It is recommended that systems for the computer-aided prediction of toxicity should be transparent and based on a thorough analysis of mechanisms of toxic action. A battery approach to the use of expert systems, in combination with suitable in vitro techniques, is likely to the most effective for overall toxicity prediction. 4. Discovering and Managing Knowledge in HTS datasets [top] Christos Nicolaou, Director, Drug Discovery Services, European Operations, Bioreason, Inc As the use of high-throughput screening systems becomes more the rule than the exception in the modern drug discovery process, the bottleneck is shifting from screening large numbers of compounds quickly and reliably, to analyzing and interpreting the massive amounts of data produced. Data analysis methods currently used by the pharmaceutical industry are almost exclusively human-driven and were designed to handle smaller and simpler datasets. These methods are quickly becoming inadequate or inefficient for thorough analysis of HTS datasets. As a result, pharmaceutical companies have recently started to invest in major improvements or replacement of their data analysis methods to be able to cope with the increased amounts and complexity of the datasets their HTS systems produce. Bioreason's automated data interpretation systems are designed to identify, characterize and prioritize lead candidates found in HTS datasets. The ultimate goal is to assist the medicinal chemist to reach a better decision in a shorter period of time regarding the quantity, the quality, the properties and thus, the value of the leads in a specific dataset. To this end, Bioreason uses a hybrid approach that combines two methods: 1. A knowledge-based system method that emulates the decision process of a pharmaceutical chemist during lead identification and development 2. Proven tools and methods from the knowledge discovery and data mining field that further explore HTS datasets by letting the data guide the search for hidden knowledge. This hybrid approach enables our reasoning systems to provide decision support for effective and efficient analysis of large amounts of screening data in a format familiar to the medicinal chemist. In addition, Bioreason's systems introduce and make use of a company-wide knowledge repository. This repository, the "knowledge-base", stores the results of the analysis of each HTS dataset in a fashion specifically designed to facilitate the reuse of the discovered knowledge during future analysis sessions or for general data mining purposes. In this presentation a description of Bioreason's vision, technology, and products will be given, along with information about the deployment of Bioreason's services in the industry. Nicolaou@bioreason.com 5. "Genetics and Computing: New Technologies To Improve Drug Discovery and Research" [top] Sheldon Ort, Information Officer for Discovery, Eli Lilly and Company My thesis today is that strategic information management for R&D in the age of pharmacogenetics represents the greatest IT challenge ever in our industry. The ability to confidently take on these high-risk, high-reward tasks may determine not only the success of your own genetic research efforts, but also your attractiveness as a partner to other organizations. The changes in information management requirements for the era of pharmacogenetics are as broad as the innovation chain itself. This is clearly seen in data points concerning known biological targets, the mapping of the complete human genome, and double-digit monthly growth rates for genetic databases. These trends present information management challenges far beyond the research lab - through clinical trials, regulatory approvals, and actually delivering the medicine to patients. A specific example can be found in the age-old quest to achieve positive medical effects without side effects. The great majority of patients who take any given drug may have no major side effects - but some do. If we can identify genetic markers for this, we can also change medicine and medical practice to improve the outcome. Obviously, there are the issues of patient privacy when DNA samples are collected. Patients and other interested parties must be convinced that they are well-protected. Furthermore, the genetic map is still a puzzle with many missing pieces, and such efforts involve massive amounts of information which must be encoded, sorted, and classified to determine what's relevant. Lilly is approaching these challenges in a variety of ways. Fundamental DNA technologies include automated genetic sequencing, constructing libraries of genetic maps and samples, and tools to display genetic information in ways that make it meaningful to scientists. We're exploring new ways of managing intellectual property issues when a scientist discovers what may be a novel protein or genetic-based target. Information systems must link genetic research across the entire discovery/development function - then across the entire enterprise and even the extended enterprise. The paradox is that the information management systems that are hardest to implement also offer the biggest payoff - true competitive advantage and great opportunities for differentiation - precisely because they are hard for competitors to duplicate.