Background RNA sequencing (RNA-Seq) is a powerful technique for transcriptome profiling of the organisms that uses the capabilities of next-generation sequencing (NGS) technologies. Recent advances in NGS let to measure the expression levels of tens to thousands of transcripts simultaneously. Using such information, developing expression-based classification algorithms is an emerging powerful method for diagnosis, disease classification and monitoring at molecular level, as well as providing potential markers of disease. Microarray based classifiers cannot be directly applied due to the discrete nature of RNA-Seq data. One way is to develop count-based classifiers, such as poisson linear discriminant analysis (PLDA) and negative binomial linear discriminant analysis (NBLDA). Other way is to transform the data hierarchically closer to microarrays and apply microarray-based classifiers. In most of the studies, the data overdispersion seems to be an another challenge in modeling RNA-Seq data. In this study, we aimed to examine the effect of dispersion parameter and classification algorithms on RNA-Seq classification. We also considered the effect of other parameters (i) sample size, (ii) number of genes, (iii) number of class, (iv) DE (differential expression) rate, (v) transformation method on classification performance. Methods We designed a comprehensive simulation study, also used two miRNA and two mRNA experimental datasets. Simulated datasets are generated from negative binomial distribution under different scenarios and real datasets are obtained from publicly available resources. We compared the results of several classifiers including PLDA with and without power transformation, NBLDA, single SVM, bagging SVM (bagSVM), classification and regression trees (CART), and random forests (RF). Results Results from the simulated and real datasets revealed that increasing the sample size, differential expression rate, number of genes and decreasing the dispersion parameter and number of groups lead to an increase in the classification accuracy. To make an overall assessment, power transformed PLDA, RF and SVM classifiers performed the highest classification accuracies. Discussion Overdispersion seems to be an important challenge in RNA-Seq classification studies. Similar with differential expression studies, classification of RNA-Seq data requires careful attention on handling data overdispersion. We conclude that, as a count-based classifier, power transformed PLDA; as a microarray based classifier vst or rlog transformed RF and SVM (bagSVM for high sample sized data) classifiers may be a good choice for classification. However, there is still a need to develop novel classifiers or transformation approaches for classification of RNA-Seq data. An R/BIOCONDUCTOR package MLSeq with a vignette is freely available at http://www.bioconductor.org/packages/2.14/bioc/html/MLSeq.html .