- Article -
| In Silico Biology 8, 0041 (2008); ©2008, Bioinformation Systems e.V. |
Effects of misdiagnosis in input data on the identification of differential expression genes in incipient Alzheimer patients
Sandeep Joseph1*, Kelly Robbins1, Wensheng Zhang1 and Romdhane Rekaya1,2,3
1 Rhodes Center for Animal and Dairy Science,
2 Department of Statistics,
3 Institute of Bioinformatics, University of Georgia, Athens, Georgia, 30602, USA
* Corresponding author
Email: sandeep@uga.edu
Edited by H. Michael; received March 27, 2008; revised September 04, 2008; accepted September 12, 2008; published October 11, 2008
Abstract
Gene expression profiles of 16 Alzheimer's (AD) patients, diagnosed as incipient or healthy using Mini-Mental State Examination and Neurofibrillary Tangles scores, were analyzed to validate the reclassification of 4 subjects previously identified as being misdiagnosed. Three datasets were created using original classifications (D1), new classifications, based on a misclassification algorithm (D2), and by removing questionable subjects (D3). Mixed model analysis was used to identify differentially expressed genes. Many genes related to the nervous system and AD were found to be differentially expressed in D2 and D3, while few genes, none related to NS or AD, were found using D1. Several additional relevant genes were found when using D2 versus D3, which were likely due to differences in sample size. These results suggest the 4 questionable subjects were likely misclassified in D1. The similarities between results obtained using D2 and D3 provides further evidence of the adequacy of the misclassification algorithm.
Keywords: gene expression, Alzheimer's Disease, misclassification algorithm, mixed model analysis, Mini-Mental Sate Examination (MMSE) and Neurofibrillary Tangles Score (NFT)