Description:

Li-Xuan Qin, PhD

Associate Member, Memorial Sloan Kettering Cancer Center, New York, NY

Presentation Title:

On the ‘Off-Label’ Uses of Data Normalization: From Microarrays to Sequencing

Abstract:

It is widely accepted that a crucial step to reproducible analysis of transcriptomics data is to identify data artifacts that arise from disparate experimental handling and remove them through data normalization. Statistical methods for data normalization were first developed for microarrays primarily in the setting of differential expression analysis. However, they have since been borrowed for other data types and analysis settings without rigorous re-evaluation. We set out to systematically assess their performance in connection with analysis settings, for microarrays and sequencing, by collecting robustly benchmarked data, devising novel resampling algorithms for simulating additional data, and developing useful criteria for performance evaluation. In this talk, I will present our findings in the settings of sample classification and survival prediction, two important analysis goals for translational biomedical research, pointing out potential caveats of popular normalization methods such as quantile normalization and providing possible solutions.

For Zoom link please contact Lucia Ontiveros : lucia.ontiveros@northwestern.edu