Ryung S Kim1, Juan Lin. 1. Department of Epidemiology and Population Health, Albert Einstein College of Medicine, Bronx, NY 10461, USA. ryung.kim@einstein.yu.edu
Abstract
MOTIVATION: Bead arrays are becoming a popular platform for high-throughput expression arrays. However, the number of the beads targeting a transcript and the variation of their intensities differ from sample to sample in these arrays. This property results in different accuracy of expression intensities of a transcript across arrays. RESULTS: We provide evidence, with publicly available spike-in data, that the false discovery rate of differential expression is reduced by modeling bead-level variability with a multi-level mixed effects model. We compare the performance of our proposed model to existing analysis methods for bead arrays: the unweighted t-test and other weighted methods. Additionally, we provide theoretical insights into when the multi-level mixed effects model outperforms other methods. Finally, we provide a software program for differential expression analysis using the multi-level mixed effects model that analyzes tens of thousands of genes efficiently. AVAILABILITY: The software program is freely available on web at http://ephpublic.aecom.yu.edu/sites/rkim/Supplementary.
MOTIVATION: Bead arrays are becoming a popular platform for high-throughput expression arrays. However, the number of the beads targeting a transcript and the variation of their intensities differ from sample to sample in these arrays. This property results in different accuracy of expression intensities of a transcript across arrays. RESULTS: We provide evidence, with publicly available spike-in data, that the false discovery rate of differential expression is reduced by modeling bead-level variability with a multi-level mixed effects model. We compare the performance of our proposed model to existing analysis methods for bead arrays: the unweighted t-test and other weighted methods. Additionally, we provide theoretical insights into when the multi-level mixed effects model outperforms other methods. Finally, we provide a software program for differential expression analysis using the multi-level mixed effects model that analyzes tens of thousands of genes efficiently. AVAILABILITY: The software program is freely available on web at http://ephpublic.aecom.yu.edu/sites/rkim/Supplementary.
Authors: Elizabeth Iorns; H James Hnatyszyn; Pearl Seo; Jennifer Clarke; Toby Ward; Marc Lippman Journal: J Natl Cancer Inst Date: 2010-07-01 Impact factor: 13.506
Authors: Kenneth Kuhn; Shawn C Baker; Eugene Chudin; Minh-Ha Lieu; Steffen Oeser; Holly Bennett; Philippe Rigault; David Barker; Timothy K McDaniel; Mark S Chee Journal: Genome Res Date: 2004-11 Impact factor: 9.043
Authors: Kristina Becanovic; Mahmoud A Pouladi; Raymond S Lim; Alexandre Kuhn; Paul Pavlidis; Ruth Luthi-Carter; Michael R Hayden; Blair R Leavitt Journal: Hum Mol Genet Date: 2010-01-20 Impact factor: 6.150
Authors: Himesh Fernando; Sven Sewitz; Jeremy Darot; Simon Tavaré; Julian Leon Huppert; Shankar Balasubramanian Journal: Nucleic Acids Res Date: 2009-09-10 Impact factor: 16.971
Authors: Mark J Dunning; Nuno L Barbosa-Morais; Andy G Lynch; Simon Tavaré; Matthew E Ritchie Journal: BMC Bioinformatics Date: 2008-02-06 Impact factor: 3.169
Authors: Douglas W Mahoney; Terry M Therneau; S Keith Anderson; Jin Jen; Jean-Pierre A Kocher; Monica M Reinholz; Edith A Perez; Jeanette E Eckel-Passow Journal: BMC Res Notes Date: 2013-01-30