A Powerful Approach for Identification of Differentially Transcribed mRNA Isoforms

Next generation sequencing is being increasingly used for transcriptome-wide analysis of differential gene expression. The primary goal in profiling expression is to identify genes or RNA isoforms differentially expressed between specific conditions. Yet, the next generation sequence-based count data are essentially different from the microarray data that are continuous type, therefore, the statistical methods developed well over the last decades cannot be applicable. For this reason, a variety of new statistical methods based on count data of transcript reads has been correspondingly developed. But currently the transcriptomic count data coming only from a few replicate libraries have high technical noise and small sample size bias, performances of these new methods are not desirable. We here developed a new statistical method specifically applicable to small sample count data called mBeta t-test for identifying differentially expressed gene or isoforms on the basis of the Beta t-test. The results obtained from simulated and real data showed that the mBeta t-test method significantly outperformed the existing statistical methods in all given scenarios. Findings of our method were validated by qRT-PCR experiments. The mBeta t-test method significantly reduced true false discoveries in differentially expressed genes or isoforms so that it had high work efficiencies in all given scenarios. In addition, the mBeta t-test method showed high stability in performance of statistical analysis and in estimation of FDR. These strongly suggests that our mBeta t-test method would offer us a creditable and reliable result of statistical analysis in practice.