PRJDB5858 Gauss-Power mixing Distributions in RNA-seq data of Arabidopsis thaliana Gene expression levels exhibit stochastic variations among genetically identical organisms under the same environmental conditions (called gene expression "noise" or phenotype "fluctuation"). In many recent transcriptome analysis based on RNA-seq, the variations of gene expression levels among replicates were empirically assumed to obey the negative binomial distribution. In this study, RNA-seq data were obtained from 8 conditions (19 ~ 26-reprlcate) of Arabidopsis thaliana, and analyzed the characteristics of gene-dependent distribution profiles of the gene expression levels. These distribution profiles could be suitably fitted by a novel distribution function named Gauss-Power mixing distribution, which was derived from a simple model of the stochastic transcription network containing the feedback loop. By such analysis, the distribution profiles of the gene expression levels were roughly divided to three types named Gaussian like type, power law like type containing long tail, and mixture type. The present fitting function predicted that the gene expression levels showing the distributions with long tail tend to be strongly influenced by the feedback of the change in their expression levels. These results also showed the features of gene expression levels correlate to their functions where the gene expression levels of essential genes tend to show Gauss type distributions while that of genes of nucleic acid binding proteins and transcription factors tend to show long tailed distributions. Gauss-Power mixing Distributions in RNA-seq data of Arabidopsis thaliana bioproject PRJDB5858 PRJDB5858 true Gene expression levels exhibit stochastic variations among genetically identical organisms under the same environmental conditions (called gene expression "noise" or phenotype "fluctuation"). In many recent transcriptome analysis based on RNA-seq, the variations of gene expression levels among replicates were empirically assumed to obey the negative binomial distribution. In this study, RNA-seq data were obtained from 8 conditions (19 ~ 26-reprlcate) of Arabidopsis thaliana, and analyzed the characteristics of gene-dependent distribution profiles of the gene expression levels. These distribution profiles could be suitably fitted by a novel distribution function named Gauss-Power mixing distribution, which was derived from a simple model of the stochastic transcription network containing the feedback loop. By such analysis, the distribution profiles of the gene expression levels were roughly divided to three types named Gaussian like type, power law like type containing long tail, and mixture type. The present fitting function predicted that the gene expression levels showing the distributions with long tail tend to be strongly influenced by the feedback of the change in their expression levels. These results also showed the features of gene expression levels correlate to their functions where the gene expression levels of essential genes tend to show Gauss type distributions while that of genes of nucleic acid binding proteins and transcription factors tend to show long tailed distributions.