Abstract Introduction RNA sequencing (RNA-seq) has rapidly become one of the main methods to study transcriptome. It has become an important and popular issue to identify biomarkers based on their differential expression patterns in NGS data. Currently, more than 10 different algorithms can be used to detect differentially expressed genes. However, challenge arises, when trying to choose the best algorithm under different experimental conditions. To address this issue, we aimed to perform a series of simulations under distinct scenarios, which can help researchers to select the best algorithm according to their data types and structures. Method RNA-Seq data were simulated by a published method named as flux-simulator. Three parameters including the proportion of differentially expressed genes, the relative fold changes of differentially expressed genes, and the replicate number of samples, were considered to define distinct simulation scenarios. To make the parameters more practical, we determined the parameters based on a real dataset. Three types of differentially expressed genes were simulated according to their fold changes between two groups. A total of 7 algorithms including DESeq, DESeq2, DEGseq, edgeR, limma, baySeq and Cuffdiff, were compared and evaluated. The raw read count table was analyzed in all algorithms, except Cuffdiff. In order to avoid errors from performing the alignment, the read count table is obtained directly from the simulated fastq file. Results Specificity and sensitivity were calculated and compared in different scenarios. As expected, the more replicate count is, the higher accuracy is obtained in all algorithms. Although previously studies showed that there were marginal effect between replicate numbers and number of reported differentially expressed genes, our results did not demonstrate such phenomenon. It might be attributed to that the replicate number is not large enough. In addition, the results showed limma, edgeR and DESeq were more conservative than the other algorithms, DEGseq has highest accuracy but follow with lower specificity when a large amount of genes are differentially expressed. Over all, edgeR shows the best trade off within sensitivity and specificity. Further research efforts are warranted to compare algorithms in different scenarios, especially the number of differentially expressed genes is low. Conclusion In conclusion, this study provided a direct comparison of different algorithms under different experimental scenarios. The results showed that edgeR algorithm worked better in the scenario of finding novel genes in new disease. With its high precision, it will be more efficiency to validate identified gene through experiment. Citation Format: Chin-Ting Wu, Mong-Hsun Tsai, Tzu-Pin Lu, Liang-Chuan Lai, Eric Y. Chuang. Performances evaluation of algorithms for identifying differentially expressed genes in RNA-seq data. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4852. doi:10.1158/1538-7445.AM2015-4852