Assessment of normalization methods in microRNA profiling data and evaluation criteria on choosing the "right" methods

• Activity Number: 582
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Hui Tang*+ and Jin Jen and Riska Shaun and Ping Yang
• Companies: Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic
• Keywords: Microarray ; Normalization ; Batch effect

Abstract

Choosing appropriate normalization methods during the data preprocessing stage in microarray-based studies can help to reduce systematic bias and infer reliable results. We presented statistical issues related to microRNA profiling data for primary lung cancer patients on a new microarray platform. We propose evaluation criteria on normalization methods, which encompasses the identification of right clusters, improving correlations between duplicated samples, increasing the number of differentially expressed genes, and reducing variations among house-keeping genes. Four batch-effect adjustment methods were compared and ranked based on each evaluation criterion. Significant gene list depends on these methods but are highly overlapping. We found that these criteria are robust and effective, and can be generalized to other microarray high throughput platforms.