"Dr. Devanarayan is the Global Head of Exploratory Statistics at AbbVie Inc. (formerly pharmaceutical division of Abbott), where his current areas of focus include biomarker & exploratory research activities from pre-clinical to late-phase and post-marketing clinical trials, patient subgroup selection using biomarkers and clinical data-mining, highthroughput-screening, genomics, proteomics, imaging, in-vivo pharmacology, and also the various assays used in GLP studies and clinical trials. He received his Ph.D. in Statistics from North Carolina State University, and has 18 years of combined pharmaceutical research experience from Eli Lilly, Merck, and AbbVie. He has filed about 10 patent applications on novel biomarker findings, given ¿ 75 invited presentations and tutorials at scientific meetings, and co-authored papers with the regulatory, academic and industry scientists on topics related to genomics, predictive modeling, clinical biomarker qualification, high throughput screening, and assays related to immunogenicity, PK and biomarker measurements. He is currently also serving in the United States Pharmacopeia (USP) Expert Panel on Immunogenicity, and is an adjunct faculty in the Division of Statistics at Northern Illinois University."

Resumo

"The importance of biomarkers in drug development has grown rapidly over the last decade. Typical uses include the establishment of proof-of-concept in early phase studies, identification of treatment responders and non-responders for personalized medicine,
prediction of safety events, etc. For these predictive applications, a single marker or a mathematical combination of multiple markers (biomarker signatures) is identified from a small targeted panel of biomarker candidates or from high-dimensional array platforms such as genomics and proteomics. Various multivariate modeling methods (e.g., logistic regression, random forests) with regularization penalties for variable selection are typically employed. For certain applications such as companion diagnostics and personalized medicine, defining a multivariate biomarker signature in terms of thresholds or cut points may be preferable. The predictive utility of these biomarker signatures should be rigorously assessed via fully embedded cross-validation methods, and further tested in independent cohorts. Other important considerations include the use of biological information from pathway analysis, batch effects, sample handling, analytical and biological variability, etc. In this seminar, an overview of the biomarker signature development process, with special emphasis on some important statistical and practical considerations, will be presented."