An effective differential expression analysis of deep-sequencing data based on the Poisson log-normal model

Tremendous amount of deep-sequencing data has unprecedentedly improved our understanding in biomedical science by digital sequence reads. To mine useful information from such data, a proper distribution for modeling all range of the count data and accurate parameter estimation are required. In this paper, we propose a method, called "DEPln," for differential expression analysis based on the Poisson log-normal (PLN) distribution with an accurate parameter estimation strategy, which aims to overcome the inconvenience in the mathematical analysis of the traditional PLN distribution. The performance of our proposed method is validated by both synthetic and real data. Experimental results indicate that our method outperforms the traditional methods in terms of the discrimination ability and results in a good tradeoff between the recall rate and the precision. Thus, our work provides a new approach for gene expression analysis and has strong potential in deep-sequencing based research.