MONDAY, SEPTEMBER 17
3:00-4:00 Conference Registration
4:00-4:10 Welcoming Remarks from Conference Director
Julia Boguslavsky, Cambridge Healthtech Institute
SYSTEMS BIOLOGY: BRIDGING THE SILOS IN
BIOMARKER DISCOVERY AND VALIDATION
4:10-4:40 Past, Present and Future of Systems Biology in Biomarker Discovery
Stephen Naylor, Ph.D., Chairman and CEO, Predictive Physiology and Medicine Inc.
Systems Biology entered the 21st Century with a roar. It offered great promise in unraveling the complexities of human biology and physiology. We all embraced its potential for providing powerful new biomarker panels with unprecedented specificity and sensitivity. The reality has been more sobering, but the potential still remains. We will review the past to learn for the future and discuss the current status of Systems Biology in Biomarker Discovery.
4:40-5:10 To be Announced
5:10-6:30 Opening Reception in the Exhibit Hall
TUESDAY, SEPTEMBER 18
7:30-8:15 Breakfast Workshop
Biological Reference Materials for Proteomics Studies
Peter Schulz-Knappe, Ph.D., Chief Scientific Officer, Proteome Sciences plc
Reference materials are routinely used in clinical chemistry to ascertain comparability of analytical tests between labs and over long time periods. The utility of proteomics is severely limited by lack of precision of qualitative and quantitative proteome analysis and poor reproducibility, even within the same lab. We produce biological reference materials for proteomics analysis by tagging plasma, serum, csf and other biological sources with isobaric mass labels, the Tandem Mass Tags (TMT®). These references are spiked into individual study samples to allow for relative quantification of the entire sample proteome with highly improved precision and CV. The use of customised references in clinical trials provides for maximum coverage of the specific proteome of disease populations.
8:30-8:35 Chairperson’s Opening Remarks
8:35-9:05 The Role of Pharmacogenomics in Discovery Translational Medicine: From Target Validation to Patient Selection
Giora Feuerstein, M.D., Assistant Vice President, Head, Discovery Translational Medicine, Wyeth Research
The complete elucidation of the human genome has opened unprecedented opportunity towards novel therapeutics, diagnostics and prognostic agents. Gene expression profiling has emerged as the primary technology that allows us to survey individuals proneness to disease (hence, disease biomarkers), response to therapeutic agents and personalized medicine. The vast progress in sequencing technology and bioinformatics suggest the possibility of commodization of gene arrays on individual levels allowing rapid medical risk assessment on a population scale. In this presentation, the role of pharmacogenomics in target validation, compound pharmacodynamic assessments, patients selection and safety monitoring will be demonstrated and reviewed.
9:05-9:35 Changing the Benefit/Risk Profile Using Pharmacogenetics
Koustubh Ranade, Ph.D., Director, Pharmacogenomics & Human Genetics, Pharmaceutical Research Institute, Bristol-Myers Squibb Company
9:35-10:05 Pharmacogenetics of Irinotecan
Jill M. Kolesar, Pharm.D., Associate Professor, Pharmacy, University of
Wisconsin - Madison
Irinotecan is a commonly used anti-cancer agent with a narrow therapeutic index. The labeling of Irinotecan was recently changed and now includes a warning of greater neutropenia risk in patients with reduced activity in the drug metabolizing enzyme UGT1A1. A known marker of reduced UGT1A1 enzyme activity is the genetic variant UGT1A1*28. In addition, the labeling now recommends clinicians consider a dose reduction in UGT1A1*28 homozygous variant patients. This labeling change has generated considerable controversy among clinicians. This talk reviews Irinotecan metabolism, Irinotecan pharmacogenetic research and the role of genetic testing prior to receiving Irinotecan therapy.
10:05-10:35 Major Pharmacogenetics Studies Using a Combination of Genome-Wide Tagging SNPs and High Density SNP Genotyping in Candidate Genes
Neil Gibson, Ph.D., Research and Development Genetics, AstraZeneca
A common goal of pharmacogenetic studies is to identify genetic biomarkers of drug efficacy and safety, which can be exploited in several ways including identifying new biological screens for drug activity and selecting suitable patients for inclusion in clinical trials. Currently there are few examples in the public domain of replicated, useful pharmacogenetic associations and predictive biomarkers. Where such markers have been reported they have generally been identified in scenarios where there was a good mechanistic understanding that enabled a hypothesis-driven approach. Here we report on examples of this type of study including the discovery of potentially useful biomarkers using genetic association where there is no strong hypothesis to guide the study design. We have performed retrospective case-control pharmacogenetic association studies using both genome-wide tagging SNP and candidate gene approaches. Typically less than 100 cases and 2-4 times as many treated controls are available for study. Up to 350,000 SNPs are selected and genotyped during the course of the investigation. Replication of initial findings is an important step to distinguish between false and true positive associations, but obtaining an independent dataset with sufficient power to test replication is challenging. However, in some instances we have been able to replicate initial findings from genome-wide studies and identify biomarkers with potential for further scientific and clinical study.
10:35-11:30 Coffee Break with Poster and Exhibit Viewing
VALIDATION OF GENOMIC BIOMARKERS
11:30-12:00 Discovery and Validation of Biomarkers Affected by Brivanib and Erbitux
J. Suso Platero, Ph.D., Group Leader, Clinical Biomarker Development, Bristol-Myers Squibb
There is an increased urgency to select patients for new treatments entering clinical trials. A way to accomplish this is by the use of biomarkers that are specific to the new compounds being tested in the clinical trials. While some of these biomarkers are being discovered in pre-clinical models, few of them have been validated for use in human subjects. One way to bridge that gap is to validate the biomarkers with technologies that are readily applicable during clinical trials. We have devised a model to discover and validate biomarkers that takes advantage of the genomic technologies and of Molecular Pathology. We make use of the new genomic technologies, like transcriptional profiling and proteomics to find biomarkers that change upon treatment with the compound in pre-clinical samples. Then, we develop assays that are amenable in Molecular Pathology. The advantage of using Molecular Pathology is that the biomarker assays can be easily translated to human assays in clinical trials. Two examples using the compounds Erbitux and Brivanib will be presented showing the discovery in pre-clinical models and its translation to clinical trials.
12:00-12:30 Bridging the Chasm Between Discovery and Validation: Breast Cancer Recurrence Case Study
John J. Sninsky, Ph.D., Vice President, Discovery Research, Celera Diagnostics
The utility of most biomarkers reported in the literature has not been confirmed in subsequent well-designed studies. FDA regulatory guidance changes as a biomarker is classified as "Exploratory", "Probable Valid" and "Known Valid." Multiple organizations and consortia have begun to address criteria for the level of validation associated with biomarker classification. Several signatures have been developed to determine risk for recurrent breast cancer based on gene expression profiles from tumor samples. The status of these signatures, relative to current biomarker classifications, will be presented.
12:30-2:00 Luncheon Technology Solutions Showcases
(Sponsorships Available. Contact Nicolas Shostak, Manager, Business Development at 781-972-5479 or
BIOMARKERS FOR CLINICAL TRIALS
2:00-2:30 Discovery and Evaluation of Predictive Response Gene Expression Signatures
Hans Winkler, Ph.D., Senior Director, Functional Genomics, Johnson and Johnson Pharmaceutical R&D
Target therapies in oncology have many advantages over chemotherapy regarding both safety and efficacy. However, due to the complex molecular make-up of tumors, the frequency of response and benefit is lower. Targeted therapies work very well but in a limited percentage of patients usually below 20%. It is therefore critical to be able to identify patients with high likelihood of benefit early during treatment and preferentially before treatment begins. We have identified a gene expression signature with the potential to identify response to a multi-targeted kinase inhibitor before treatment. The challenges in identification and, specifically, validation of such signatures will be discussed.
2:30-3:00 Correlation of Biomarker Response in Gene Expression and GTP Pools with In Vitro Cell Resistance in Patient Samples from a Phase I Trial of AVN944J
Michael Hamilton, M.D., Chief Medical Officer, Avalon Pharmaceuticals
AVN944 is an orally bioavailable inhibitor of inosine monophosphate dehydrogenase 1 and 2 (IMPDH). AVN944 is in a Phase I trial in patients with advanced AML, CLL, and Myeloma. In addition to the objective of defining the MTD and pharmacology of AVN944 (given bid for 21 days on a 28 day cycle), we have examined a comprehensive set of pharmacodynamic biomarkers in peripheral blood samples from all patients in this trial pre and post-AVN944 dosing. We see a dynamic response of subsets of genes depending on (1) the type of malignancy, (2) the predominant cell type in the peripheral blood, and (3) the dose of AVN-944 received related directly to cellular pathways and functions of IMPDH. Increasing doses resulted in increases in the number of genes changing per individual, and/or an increase in the magnitude of change in a gene-dependent manner. These genetic markers, correlated with biochemical effects of the drug on protein function and guanine nucleotide levels, can be used as important biomarkers in selecting the optimal dose for additional studies with this drug and for enriching for patients most likely to respond to the drug. Further, these biomarkers serve as surrogates for monitoring clinical
activity of the drug in ongoing trials.
3:00-4:00 Refreshment Break with Poster and Exhibit Viewing
BIOMARKER DATA ANALYSIS
4:00-4:30 Comparing Fingerprint-based and Biomarker-based Classifiers
Brian Luke, Ph.D., Senior Scientist, Advanced Biomedical Computing Center, National Cancer Institute
There are two different definitions of biomarkers that, when used in an appropriate classifier, are used to determine whether or not an individual has a particular disease. The first is a panel of markers that are used to construct fingerprints to determine the health category of an individual. Examples of published algorithms include a decision tree (DT), an artificial neural network (ANN), and the medoid classification algorithm (MCA) as used by the laboratories of Perticoin and Liotta. The second is a biomolecule whose full range of concentrations is strongly controlled by the individual’s biochemical state. It has been argued that if a classifier can be constructed from a training set that is able to accurately classify a blinded testing set, then that classifier must be based on some underlying biological principle. This talk will show that the concept that we call coverage strongly affects fingerprint-based classifiers when the available data are divided into training and testing sets. We show that the application of DT or MCA classifiers on datasets that contain random peak intensities are able to accurately classify properly selected training and testing sets even though there is no biological relevance. In contrast, a biomarker-based classifier is unable to construct an accurate classifier using the same random datasets. The inference of biological information from an accurate classifier is only true for a biomarker-based classifier since fingerprint-based classifiers are able to accurately classify a dataset by chance.
4:30-5:00 Discovering Multivariate Biomarkers and Informative Sets of Genes
Darius Dziuda, Ph.D., Assistant Professor, Mathematics, Central Connecticut State University
The presentation will outline and discuss current issues in biomarker discovery, especially multivariate approach to identification of genomic biomarkers for medical diagnosis, prognosis, and drug discovery. In the second part of this talk, a novel method associating a small multivariate biomarker with a larger Informative Set of Genes will be presented. The Informative Set of Genes is important for elucidation of biological processes underlying class differentiation.
5:00-5:30 Chemosensitivity Biomarker Extrapolation from NCI-60 Cancer Cell Lines to Clinical Tumors for Predicting Patients’ Chemotherapeutic Responses
Jae K. Lee, Ph.D., Associate Professor, Biostatistics, University of Virginia School of Medicine; and Director, UVA Bioinformatics Support Core
The U.S. National Cancer Institute has used a panel of 60 diverse human cancer cell lines (the NCI-60) to screen >100,000 chemical compounds for anticancer activity. However, not all important cancer types are included on the panel nor are drug responses on the panel predictive of clinical efficacy in patients. We thus asked whether it would be possible to identify common chemosensitivity biomarkers from that rich database to predict drug activity in cell types not included in the NCI-60 panel or, even further, clinical responses in patients with tumors. We address that challenge by developing a novel algorithm “Co-eXpression ExtrapolatioN” (COXEN), which can effectively identify concordant genomic biomarkers between two independent expression profiling data sets, here extrapolating the genomic expression patterns of NCI-60 chemosensitivity biomakers with those of clinical tumors. Applying our COXEN approach in a prospective fashion, we predicted anticancer drug activities on completely independent bladder cancer, which is not included in the NCI-60 panel, and on breast cancer patients treated with commonly used chemotherapeutics with >80% accuracy. We also used COXEN for in silico screening of 45,545 compounds and identified a novel agent with superior growth inhibition activity against human bladder cancer.
5:30-6:00 Decipher and Select the Genomic Biomarkers with the DAVID Bioinformatics Resources
Richard A. Lempicki, Ph.D., LIB Lab Head, Senior Scientist, Clinical Services Program, Laboratory of Immunopathogenesis and Bioinformatics, SAIC-Frederick, Inc.
In the post-genomic era, one of the challenges is to systematically and comprehensively interpret large amounts of “omic” data results, such as biomarker discovery based on microarray studies. Using the biological knowledge accumulated in the past decades and the aid of computing algorithms, it is possible to assemble potential biological pictures associated with these studies toward the more specific biomarker discovery. The DAVID (The Database for Annotation, Visualization and Integrated Discovery) Bioinformatics Resources provides a set of powerful, novel tools that researchers can use to explore their large gene lists in-depth from many different biological angles in order to extract associated biological meanings to the greatest extent possible. For example, the DAVID Gene Functional Classification Tool is able to quickly group large gene lists into functional gene groups. Therefore, it increases the likelihood that the investigator will identify more specific biomarkers most pertinent to the biological phenomena under study. To present the usefulness of the DAVID Bioinformatics Resources in biomarker discovery, a real life clinical biomarker study conducted in my laboratory will be presented. In this study, with the help of the DAVID Bioinformatics Resources, a set of interferon-stimulated genes, as clinical biomarkers, were successfully selected from Affy microarray studies on the HCV/HIV co-infection patients. The marker genes are able to predict the output of anti-HCV therapy for the HCV/HIV co-infected patients.