Combined With The Control Information Of The Genome Microarray Data Analysis

DNA microarray experiments are usually classified based on the type of array(cDNA and oligonudeotide) or according to the sample(static and time-series experiments).In static expression experiments,a snapshot of the expression of genes in different samples is measured,while in time series expression,a temporal process is measured.Many statistical methods have been established to analyze these different kinds of microarray datasets. However,individual microarray experiments are often restricted by the amount of samples and lack of repeats,so the obtained results are correspondingly short of conviction.Using standardized microarray database such as GEO and ArrayExpress,we handpicked the microarray datasets about same biological objective from different laboratories,and carried out a meta-analysis on them.Fisher formula and permutation test are the core steps during the process of our meta-analysis.For the case study in static microarray data,we chose the rising microRNA expression profiles in human prostate cancer.At statistic generated by randomly assigning the sample lables to expression values of the miRNA.The P value then equaled the fraction of random t statistics that were great than or equal to the actual t statistic.These Ps were using for calculating combined S value through Fisher formula.Resampling the Ps to generate random S value,we could give each miRNA a combined significance. After FDR rectify,Q value can be obtained for each miRNA.For the time-series datasets,we implemented the same procedure of permutation and combation,except introducing the S area value for describing each gene's expression change along time.Furthermore,in the case study of yeast transient heat shock microarray datasets,we established a comprehensive strategy for meta-analysis of time-series array data which having short time points and lacking of repeats.Integrating with the regulatory and other biological information is necessary for systematic study of microarray datasets.We downloaded the microRNA target gene lists and investigated their targets' transcript and protein sysnthesis levels in prostate cancer.An interesting relationship between miRNA and their targets' levels in prostate cancer has been found;as for the dynamic yeast heat shock datasets,we also found some regulatory tendencies between Hsf1 and Msn2/4 system through clustering and promoter analysis.With the increasing number of publicly available datasets,there are many opportunities and challenges to incorporatively analyze their results.Coupled with gene regulatory information,we hope that,the strategy for array meta-analysis established by us could help researchers to recycle the extensive datasets generated by different groups and,could mine some hidden and valuable biological information.