Optimal Power Transformations for TMA Biomarker Data

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Bhupendra Rawal*+ and Daohai Yu and Michael J. Schell
• Companies: Moffitt Cancer Center and Research Institute and Moffitt Cancer Center and Research Institute and Moffitt Cancer Center and Research Institute
• Keywords: Transformation ; Biomarker

Abstract

Pre-processing data using raw (R), log (L), square-root (S), or quarter-root (Q) as an optimal transformation has been accepted in observational and clinical studies. There is a special need to look for an optimal transformation in tissue-microarray (TMA) data. 53 TMA runs of biomarker expression from 187 early stage non-small cell lung cancer patients were studied to assess the optimal power transformations. The automated-quantitative analysis (AQUA) was used to measure the nucleus and cytoplasmic scores. The goal of optimal transformation is to homogenize the variability of AQUA scores in order to minimize the influence of any individual score(s). We used the MM robust regression method for estimation and selected as the optimal transformation the one with the lowest average number of high leverage points. 7 (13%) runs favor R for analysis and the other 46 (87%) runs require a transformation. Among the 46 runs, 13 (25%), 13 (25%), and 10 (19%) favor L, S and Q, respectively, while 8 (15%) runs select either L or Q, and 2 (4%) runs either S or Q. Thus, pre-processing biomarker data to determine an optimal transformation is an important step in the statistical analysis of TMA data.