碩士 / 國立陽明大學 / 公共衛生研究所 / 96 / Introduction:
Most complex diseases are controlled by gene(s) or interaction between gene(s) and environmental factors. Several methods have been proposed to detect interaction between trait-associated genes or loci. Multifactor dimensionality reduction (MDR) and logistic regression (LR) are two commonly used methods to target this question. However, there is no complete comparison between these two methods. In this study, we considered different parameters (including disease models, allele frequency, sample sizes and penetrance) to compare the performance between MDR and LR in the presence of gene-gene interaction.
Methods:
Under the assumption of existing gene-gene interaction, we set up five different two loci disease models, sample sizes, allele frequency and penetrance. We simulated 1000 data set for each model. For each data set, equal numbers of case and control subjects as well as their genotype data were generated by the SNaP software. We used the SAS package for stepwise logistic regression analyses and the MDR software for MDR analyses. Finally, we compared power, false discovery rate and false negative rate between these two methods.
Results:
For all models and parameter combinations, MDR always have better power than logistic regression. Under the majority situations, the false negative rate of MDR are lower than logistic regression. Moreover, the false discovery rate of MDR are lower than logistic regression under low penetrance, but the false discovery rate of MDR are higher than logistic regression under high penetrance. However, the false negative rates of two methods are decreased if the sample sizes are increased, but the false discovery rate may increase as well.
Conclusion:
Our results suggest that the MDR method is always performed better than the logistic regression to detect gene-gene interaction for complex disease, however, the MDR method might have higher probability of false discovery.
| Identifer | oai:union.ndltd.org:TW/096YM005058029 |
| Date | January 2008 |
| Creators | Pei-Chi Liu, 劉佩綺 |
| Contributors | Ming-Wei Lin, 林明薇 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 63 |
Page generated in 0.0127 seconds