Statistical analysis of GeneChip microarrays for differential expression and biomarker identification

Presented: August 3, 2005

Speaker: Michael O'Connell, Insightful Corporation

Listen to the archived Web cast.

Download the Web cast presentation [PDF]

Abstract: This educational seminar describes workflows for statistical analysis of microarray GeneChip® data in combination with clinical (endpoint) data with a focus on the identification of biomarkers. Biomarkers can enhance the understanding of a drug's mechanism-of-action and/or toxicity profile and may be used to demonstrate clinical dose-response, define dosing frequency, dose range and study design of early phase clinical trials including subject identification and/or stratification.

Particular attention will be given to the analysis of two-sample and multi-level phenotype expression data and the workflows described include careful attention to all facets of microarray/clinical data analysis including pre-processing (normalization, probe-level summary), differential expression testing (multi-factor linear models), class discovery (clustering), class prediction (machine learning), gene-list management and annotation (pathway connections). All of these workflow steps are important in the accurate identification of biomarkers.

The workflows and analyses will be described primarily from the perspective of the end-user scientist. Some examples of current work will be presented from the perspective of the analyst and computational biologist e.g. (a) feature selection and filtering using tree ensembles and (recursive) support vector machines, and (b) survival analysis using clinical endpoints as the model response variable and gene expression as the explanatory variables.

The analyses use combinations of S-PLUS and S+ArrayAnalyzer, the microarray data analysis software made by Insightful. Deployment of statistical analyses through a web user interface and a Spotfire user interface will be discussed. Analyses in exploratory and regulated (e.g. 21 CFR 11) environments will be described.

References:

• S+ArrayAnalyzer datasheet
• Clinical Development Web site
• Insightful Research Libraries
• Case studies in S-PLUS: Millard and Krause
• S-PLUS books
• S+Graphlets Info

Michael O'Connell, Ph.D., Insightful Corporation

Michael O'Connell, Ph.D., is director of Life Science Solutions at Insightful Corp. He has more than 15 years experience in the medical device, informatics and health-care statistics arena, having published more than 30 papers on statistics, data mining and health-care applications. This has included statistical methods work in the areas of non-parametric regression, experimental design, calibration and mixed models; and applications such as DNA amplification, diagnostics, microarray data analysis and safety data mining.