Click here to view a post conference review and photos

Immediately followed by CHI's Second Annual Data Visualization and Interpretation
and Data Integration for the Pharmaceutical Industry.

Corporate Sponsors:

Scientific Advisors
Mr. Thomas J. Downey, Partek, Inc.
Dr. Sorin Draghici, Wayne State University
Dr. Robert Nadon, McGill University
Dr. Hrissi Samartzidou, Amersham Biosciences
Dr. Jay P. Tiesman, Procter & Gamble

Speakers
Dr. Amber Anderson, GlaxoSmithKline
Dr. Gabor Balazsi, Northwestern University Feinberg School of Medicine
Dr. Eric Bremer, Children's Memorial Hospital, Chicago
Dr. Hyungjun Cho, University of Virginia
Dr. Sorin Draghici, Wayne State University
Dr. Greg Gibson, North Carolina State University
Dr. Rafael A Irizarry, Johns Hopkins University
Dr. Michael O'Connell, Insightful Corporation
Dr. Jack Pollard, Jr., 3rd Millennium
Dr. Petra Ross-Macdonald, Bristol-Myers Squibb
Dr. Hrissi Samartzidou, Amersham Biosciences
Dr. Richard Simon, National Cancer Institute
Dr. Edward Spitznagel, Washington University
Dr. Phillip Stafford, BioMining
Dr. Christian Stratowa, Boehringer-Ingelheim Austria
Dr. Kenton Juhlin, Principal Scientist, Biometrics and Statistical Sciences
Dr. Yuhai Tu, IBM T. J. Watson Research Center
Dr. Youxiang Wang, Full Moon bioSystems
Dr. Lee Weng, Rosetta Biosoftware
Dr. Jun Zou, Schering-Plough Research Institute

Experimental Design
Design of DNA Microrarray Studies
Using Variance Component Information
Taking an Integrated Approach
Managing Experimental Noise
Quantitative Noise Characterization

Data Quality
Probe Arrangement
Ensuring Reliable Microarray Data
Expression Calibration Standard
Microarray Platform Comparison
Storage and Analysis of Microarray Data in the Terabyte Range

Data Assessment
GeneChip Probe Level Analysis
Advantages of Error-Model Based Analysis
Bayesian Hierarchical Error Model
Minimizing False Discovery and Maximizing Power in Small Microarray Experiments
Data Analysis in the Post-Genomic Era

Applications of Data Analysis
Monkey Model of Allergic Asthma
Genetic Changes in NSAIDs versus Placebo
Genotype-by-Treatment Interactions
Compound Selection
Classification of Pediatric Brain Tumors

PROGRAM

Sunday, September 21

5:00-6:00 Early Conference Registration

Monday, September 22

7:30 am Registration and Coffee

Experimental Design

8:30 Chair's Opening Remarks
Mr. Thomas J. Downey

8:40 Design of DNA Microarray Studies
Dr. Richard Simon, Chief, Biometric Research Branch; Head, Molecular Statistics & Bioinformatics, Section, National Cancer Institute
DNA microarray experiments require planning. Planning is driven by the experimental objectives. Good DNA microarray experiments have clear objectives. The objectives are generally not based on gene-specific mechanistic hypotheses, but it is erroneous to conceive of DNA microarray investigations as aimless data mining in search of unanticipated patterns that will provide answers to unasked questions. I will present an evaluation of hybridization designs for DNA microarray studies for class comparison and class discovery objectives. Much of the conventional wisdom about DNA microarray study design is not well founded. Designs I will cover include common reference designs, balanced block designs, loop designs, and dye swap designs. Other topics covered include limitations of RNA pooling, relative merits of biological versus technical replication, and number of samples required.

9:10 Using Variance Component Information in Designing Microarray Experiments
Dr. Edward Spitznagel, Professor of Mathematics and Biostatistics, Washington University
Variation in microarray data is due to many different sources. By blocking over the major sources of variation, it is possible to reduce the error variance and thereby improve statistical power. Information obtained from existing experiments allows us to estimate the variance components of the major sources of variation. Use of this information for experimental design and sample size determination will be illustrated with concrete examples.

9:40 Taking an Integrated Approach to Experimental Design and Data Analysis
Dr. Jack Pollard, Jr., Director of Informatics Research, 3rd Millennium
We will discuss our recent SBIR-funded work with researchers at the Department of Defense to deploy an informatics resource for the DOD's Biomedical Technology Area program in Infectious Disease of Military Importance. They are studying the interactions of HIV and malaria. We will outline an integrated approach to experimental design and data analysis that is embodied in a unique system that allows researchers to analyze their expression data in a statistically rigorous fashion by leveraging the design of the experiment, the biology of the samples, and the information recorded in a LIMS. We will discuss the functional requirements and design considerations of such a system as well as the results of the integrated analysis.

10:10 Poster and Exhibit Viewing, Refreshment Break

11:00 Assessing and Managing Experimental Noise from Small-Sample Microarray Experiments
Dr. Kenton Juhlin, Principal Scientist, Biometrics and Statistical Sciences
Under normal conditions, there are myriad sources of variability in microarray experiments. These must be managed through careful experimental design in order to optimize data quality. As these technologies mature, we are placing even greater demands upon them, including the acquisition of high-quality data from smaller and smaller samples. These demands introduce additional sources of experimental variability that also must be identified and addressed. We will describe our continuing efforts to improve the reliability of experiments using very small quantities of input RNA.

11:30 Quantitative Noise Characterization and Physical Modeling of Gene Expression Microarray Data
Dr. Yuhai Tu, Physical Sciences & Computational Biology Center, IBM
T. J. Watson Research Center Noise is a big problem in analyzing gene expression data. Together with our experimentalist colleagues at Columbia University, we have designed and carried out a set of replicate experiments, which bifurcate at different stages of the experimental procedures. Through analyzing the replicate data, we were able to separate the contributions to the total noise from sample preparation and the actual hybridization process, and obtained the quantitative noise characterization for each of the independent noise sources, in particular their dependence on the expression level itself. In a related work, we have constructed a physical model based on hybridization kinetics and the probe sequence information to understand the relation between the microarray readout and the true mRNA concentration. Our study shows strong correlation between hybridization energy and the microarray data, and we propose a algorithm to obtain the true mRNA concentration from the hybridization intensity measured by gene expression microarray.

12:00 Panel Discussion

Data Quality

2:00 Chair's Remarks
Dr. Kenton Juhlin

2:05 Log-Ratios of Gene Expression Depends on the Arrangement of Probes on the Microarray
Dr. Gabor Balazsi, Research Associate, Pathology, Northwestern University Feinberg School of Medicine
Global transcriptome data is increasingly combined with sophisticated mathematical analyses to extract information about a cell's functional state. Yet the extent to which the results reflect experimental bias at the expense of true biological information remains largely unknown. We show that the spatial arrangement of probes on microarrays and the particulars of the printing procedure significantly affect the log-ratio data of mRNA expression levels measured during the Saccharomyces cerevisiae cell cycle. We present a numerical method that filters out these technology-derived contributions from the existing microarray data and discuss the possiblity to diminish this effect in future experiments by improved microarray design.

2:35 CodeLink™ Expression Bioarrays: Ensuring Reliable Microarray Data
Dr. Hrissi Samartzidou, Senior Product Manager, Amersham Biosciences
The usefulness of DNA microarrays for gene expression profiling depends on the quality of the microarray results. Sensitivity, specificity, precision, and accuracy are all important for analysis of complex expression profiles using microarray technology. We have heavily invested in the reduction of variability observed in microarray experiments and the reliable performance of microarrays. The techniques employed for quality control and to assess the performance of the CodeLink System will be described, including design and implementation of highly specific probes, manufacturing controls, system performance metrics, and statistical methods for determining the reliability of the microarray data. Examples of the CodeLink Expression System performance will be used to describe how high-quality experimental results lead to better understanding of biological systems.

3:05 Standardization of Microarray Data with Expression Calibration Standard
Dr. Youxiang Wang, Chief Executive Officer, Full Moon bioSystems
The reproducibility and reliability of quantitative microarray results are not sufficiently robust to permit widespread application of this technology in diagnostic and screening areas. Therefore standards are urgently needed. This presentation will address two issues for standardization microarray data: the quantitation of fluorescence using microarray readers and the quantitation of biological samples.

3:35 Poster and Exhibit Viewing, Refreshment Break

4:15 Microarray Platform Comparison
Dr. Phillip Stafford, Chief Executive Officer, BioMining
Expression microarrays have proven invaluable for identifying patterns of expression among groups of genes. Unfortunately, cross-platform validation has been difficult because of the proprietary nature of the data, the expense of running identical experiments, and the difficulty imposed by multiple experimental sites. We present 4 leading microarray platforms hybridized to human heart and liver, and the rather surprising results.

4:45 A Novel Framework for Distributed Storage and Analysis of Microarray Data in the Terabyte Range: An Alternative to BioConductor
Dr. Christian Stratowa, Bioinformatics Scientist, NCE Lead Discovery, Boehringer-Ingelheim Austria
Novel high-throughput technologies such as DNA microarray analyses are allowing biologists to generate sets of data in the terabyte realm. Much of this data will be deposited in the public domain, necessitating a common standard. Currently available database systems are not appropriate for these intentions. As an alternative, I will introduce ROOT, an object-oriented framework that has been developed at CERN for distributed data warehousing and data mining of particle data in the petabyte range. Data is stored as sets of objects in machine-independent files, and specialized storage methods are used to get direct access to separate attributes of selected data objects. ROOT has been designed in such a way that it can query its databases in parallel on SMP/MPP machines, on clusters of PC's, or using common GRID services. In order to demonstrate the applicability of ROOT to microarray data, I will present a functional prototype system, called XPS - eXpression Profiling System, which can be considered to be an alternative to the BioConductor project. The current implementation handles efficient storage of Affymetrix GeneChip schemes, gene annotation and data, and the pre-processing, normalization and filtering of GeneChip data. Based on ROOT and XPS, I will propose a novel standard for the distributed storage of microarray data.

5:15 Panel Discussion

5:45-6:30 Networking Reception
Pizza and "Micro"brews (sponsored by Cambridge Healthtech Institute)

Tuesday, September 23

Data Assessment

8:30 Chair's Remarks
Dr. Sorin Draghici, Assistant Professor, Department of Computer Science, Wayne State University

8:35 New Developments In Affymetrix GeneChip Probe Level Data Analysis
Dr. Rafael A Irizarry, Assistant Professor, Department of Biostatistics, Bloomberg School of Public Health, Johns Hopkins University
This presentation expands some of the work that lead to the popular alternative to Affymetrix's MAS 5.0 expression measure: the Robust Multi-Array Analysis (RMA) by examining the possibility of using sequence information to improve normalization and background correction methods. The description of the preliminary work on a model-based approach that unifies the background correction, normalization, and summary statistic steps are highlighted.

9:05 Advantages of Error-Model Based Microarray Data Analysis
Dr. Lee Weng, Director of Applied Research, Rosetta Biosoftware
The increasing volume and complexity of microarray data have created demand for more powerful analysis tools. These new tools are developed with the primary goals of achieving higher statistical analysis power and discovering more valuable information in the data. Although the total amount of data is enormous in microarray studies, the number of replications is usually too small to make reliable statistical inferences based on conventional statistical methods. The error model technology in the Rosetta Resolver system provides a significant benefit in analyzing microarray data with few replications. It provides significantly higher sensitivity and specificity in statistical analyses, such as t-tests and ANOVA, than analysis methods that do not leverage error models.

9:35 Bayesian Hierarchical Error Model for Analysis of Gene Expression Data
Dr. Hyungjun Cho, Research Associate, Health Evaluation Sciences, University of Virginia
Analysis of genome-wide microarray data requires estimation of a large number of genetic parameters of individual genes and their interaction expression patterns under multiple biological conditions. The widely used ANOVA/linear modeling approaches (e.g. Kerr and Churchill, 2001) are limited because only a single common-error factor can be characterized for all genes and varieties in the model, whereas different genes have quite heterogeneous error variability and the sources of microarray error variability comprise various biological and array-experimental factors, such as individual replication, sample preparation, hybridization, and image processing. In order to overcome this limitation, we have developed a Bayesian hierarchical error (HEM) modeling method by MCMC (Markov Chain Monte Carlo) estimation with its efficient, convenient software available at our web site.

10:05 Poster and Exhibit Viewing, Refreshment Break

10:45 Statistical Issues in the Analysis of Microarray Data: Minimizing False Discovery and Maximizing Power in Small Microarray Experiments
Dr. Michael O'Connell, Director, BioPharmaceutical Solutions, Insightful Corporation
Since microarrays are expensive, and target RNA sample available is often limited, experiments are typically done with a limited number of replicates. Several statistical testing methods for differential gene expression have been suggested, but some of these approaches are underpowered and have high false-positive rates because within-gene variance estimates are based on a small number of replicated arrays. This talk describes methods which borrow strength from the set of genes on the chip in the construction of the error estimates used in signal-to-noise ratios. These methods provide better control of the false discovery rate and false negative rate, and handle the situation where a gene may have very similar expression values in duplicate chips by chance alone. Such situations result in false positives using a within-gene error model.

11:15 Onto-Tools: Data Analysis in the Post-Genomic Era
Dr. Sorin Draghici
Independently of the methods used to select the genes regulated in a given experiment, the common task faced by any researcher is to translate these lists of genes into a better understanding of the biological phenomena involved. We developed Onto-Express (OE) as a novel tool able to automatically translate such lists of differentially regulated genes into functional profiles characterizing the impact of the condition studied. Onto-Express is complemented by another set of tools including Onto-Compare, Onto-Design and Onto-Convert. Onto-Compare allows the user to perform a functional comparison of any number of commercially available microarrays. If none of the commercially available tools is found to be suitable, Onto-Design allows researchers to design their own custom array by selecting the best set of genes representing given selected pathways and biological processes. Onto-Convert allows a quick translation between UniGene cluster IDs, NCBI accession IDs and custom, array specific IDs such as Affymetrix IDs.

11:45 Panel Discussion

Applications of Data Analysis

1:30 Chair's Remarks
Dr. Hrissi Samartzidou

1:35 Microarray Profile of Differentially Expressed Genes in a Monkey Model of Allergic Asthma
Dr. Jun Zou, Principal Scientist, Allergy Research, Schering-Plough Research Institute
Inhalation of the antigen Ascaris suum by allergic monkeys causes an immediate bronchoconstriction and delayed allergic reaction. To identify genes involved in this process, the gene expression pattern of allergic monkey lungs was profiled by microarrays. Monkeys were challenged by inhalation with Ascaris suum antigen or IL-4; lung tissue was collected at 4, 18, or 24 hour after antigen challenge. Each challenged monkey lung was compared to a pool of normal, unchallenged monkey lungs. Of ~40,000 cDNA elements on microarray tested, 169 were identified to be regulated by > 2.5-fold. Cluster analysis revealed at least 5 groups of genes with unique expression patterns. Confirmation of differential expression of selected genes was obtained for 95% in the original tissue using real-time PCR.

2:05 Analysis of Microarray Data from Healthy Volunteer Clinical Trial to Investigate Genetic Changes in NSAIDs versus Placebo
Dr. Amber Anderson, Senior Statistician, Clinical Pharmacology Statistics and Programming, GlaxoSmithKline
We will present the statistical analysis methods and results of a crossover clinical trial whose major objective was to investigate differences in genetic expression of healthy volunteers after single dosing and steady-state dosing of marketed NSAIDs versus placebo. Statistical analysis of the data involved fitting an analysis of variance model for each gene separately, incorporating factors appropriate to the study design, and implementing a permutation method to adjust for multiple comparisons (ie., multiple genes). This analysis generated a list of the most significantly differentially expressed genes for each comparison which is being investigated to link these identified genes and known genetic and NSAID functions.

2:35 Quantification of Genotype-by-Treatment Interactions using the SAS Microarray Solution
Dr. Greg Gibson, Associate Professor, Genetics, North Carolina State Univ.
We will describe how mixed model analysis of variance implemented in the SAS Microarray Solution can be used to tease apart contributions of multiple interacting factors to expression profile variation. Moderate replication gives sufficient power to detect highly significant 1.3 fold differences among genotypes, sexes, drug responses, and other environmental factors. Examples from Drosophila and other organisms highlight ways that transcription profiling can supplement genotype-based association mapping in the dissection of complex traits.

3:05 Poster and Exhibit Viewing; Refreshments and Desserts Served

3:45 Expression Profiling in Support of Compound Selection
Dr. Petra Ross-Macdonald, Senior Research Investigator, Applied Genomics, Bristol-Myers Squibb
Even after extensive triage through a broad range of assays, a drug discovery program may be left with dozens of compounds that pass all criteria yet have hidden liabilities of selectivity that can be catastrophic. We are analyzing the gene expression profiles of treated cells to uncover such differences, providing a novel means of characterizing and prioritizing compounds. We will discuss early implementation of this approach on a kinase inhibitor program.

4:15 Molecular Classification of Pediatric Brain Tumors by Gene Expression Profiling
Dr. Eric Bremer, Director, Brain Tumor Research, Neurosurgery, Children's Memorial Hospital, Chicago
Pediatric brain tumors are the most common type of solid tumor in children. We have built predictive models for the classification of 6 types of pediatric brain tumors including sub-types to aid in the diagnosis of these tumors. This was accomplished using artificial neural networks and decesion trees on data orginating from different institutions. The accuracy of these models were found to be between 90 and 95% on independent validation sets. These data indicate that gene expression analysis can provide objective tool for the clinical diagnosis of brain tumors.

4:45 Panel Discussion

5:15 Close of Conference

Wednesday, September 24

POST-Conference Short Course Tutorials

Call for Posters:
Cambridge Healthtech Institute encourages attendees to gain further exposure by presenting their work in the poster sessions. Please fill out the registration form, with the poster title and primary author. To ensure inclusion in the conference CD, a one-page ELECTRONIC abstract must be submitted and registration must be paid in full by August 22, 2003.


Gene Logic provides a broad range of genomics information and bioinformatics solutions and integrated contract research services to the pharmaceutical and biotechnology industry. Through established expertise in biosample collection, handling and processing, genomic data production, data management and software systems development, and the integration and completion of preclinical and clinical studies and regulatory submissions, the Company's information products and research services facilitate and expedite the drug discovery and development process.

Lead Publication:
Sponsoring Publications:
Web Partner:

There are many sponsorship opportunities for your company to maximize its exposure and influence. They include conference-specific sponsorships, technology workshops, networking receptions, delegate bags, etc. We are also ready to work with you in customizing a solution to meet your specific marketing objectives. Make a lasting impression by taking advantage of these marketing tools. For exhibit and sponsorship information, please contact Carol Dinerstein at 781-972-5471 or dinerstein@healthtech.com.

TRAVEL INFORMATION
Special Airline Discounts Available
Special Zone and Discount Fares have been established for this conference with United Airlines. Please call United Airlines Meeting Reservation Desk at 800-521-4041 and reference ID#579YS.

HOTEL INFORMATION
Wyndham Baltimore Inner Harbor
101 W. Fayette Street
Baltimore, Maryland 21201
T: 410-752-1100 o F: 410-752-0832
Cut-off date: August 29, 2003
$179 single/$199 double occupancy

Please call the hotel directly to make your room reservation. Identify yourself as a Cambridge Healthtech Institute conference attendee to receive the reduced room rate. Reservations made after the cut-off date or after the group room block has been filled (whichever comes first) will be accepted on a space-and-rate-availability basis. Rooms are limited, so please book early.