Recently after human genome sequence has been determined almost perfectly, more and more researchers have been studying genes in detail. Therefore, we are sure that accumulated gene information for human will be getting more important in the near future to develop customized medicine and to make gene interactions clear. Among plenty of information, micro array might be one of the most important analysis method for genes because it is the technique that can get big amount of the gene expressions data from one time experiment and also can be used for DNA isolation. To get the novel knowledge from micro array data, we need to enrich statistical tools for its data analysis. So far, many mathematical theories and definition have been proposing. However, many of those proposals are tested with strict conditions or customized to data for specific species. In this paper, we reviewed existing typical statistical methods for micro array analysis and discussed the repeatability of the analysis, construction the guideline with more general, procedure. First we analyzed the micro array data for TG rats, with statistical methods of family-wise error rate (FWER) control approach and False Discovery Rate (FDR) control approach. As existing report, no significantly different gene could be detected with FWER control approach. On the other hand, we could find several genes significantly with FDR control approach even q=0.5. To find out the reliability of FDR control approach with micro array conditions, we have analyzed 2 more pieces of data from Gene Expression Omnibus (GEO) public database on the web site with SAM in addition to FWER and FDR control approaches. We could find a certain number of significantly different genes with BH method and SAM in the case of q=0.05. However, we have to note that the number and kinds of detected genes are different when we compare our result with the one from the published paper. Even if the same approach is used to analyze the same micro array data, we might get a different result because the distinct definition for micro array data has not been set yet. It means that from the same data we will get different results depending on researchers. We are afraid that this problem will have a big effect on developing new medicines and to progress the next step, like a 2~(nd) screening. So, we suggest that we should have certain guidelines to analyze Micro-Array data validly with statistic method and it will surely be helpful for Micro-Array analysis for medical studies in the future.
展开▼