Non-parametric MANOVA methods for detecting differentially expressed genes in RT-PCR experiments

RT-PCR is a quantitative technique of molecular biology used to amplify DNA sequences starting from a sample of mRNA. RT-PCR is typically used to explore gene expression variation across groups of treatment. Because of the non-normal distribution of data, various authors have proposed nonparametric methods based on the MANOVA approach to deal with this problem. These methods, based on the use of distances (Anderson et al.) and of the use of the medians instead of means (Xu et al.) allow to explore patterns of differential gene expression via global F-ratio test obtained through permutations and subsequent analysis of contrasts. A non-parametric test for multiple univariate comparisons is then used to evaluate which genes are likely to be differentially expressed across groups. Results of a study involving 30 mice assigned to one control group and to 4 different treatments are presented. An effect of treatment on gene expression is detected by both methods, with good agreement also with respect to the pre-selected multiple comparisons. These results are potentially useful to draw out new biological hypothesis to be verified in following designed studies. Future research will be focused on comparison of such methods with classical strategies for analysing RT-PCR data, to better evaluate their advantages and drawbacks; moreover, work will also concentrate on extending such methods to doubly multivariate design, a context often arising in biological data.