The Cross-Entropy method and multiple change-points detection in zero-inflated DNA read count data

We model DNA count data as a multiple change point problem. This means that the data are divided into multiple segments based on the unknown number of change points. Each segment of the process is modeled by using a zero modified count data distribution. We observe that zeroinflated negative binomial (ZINB) model fits the data better than the competing negative binomial (NB) model. We propose an extension to the Cross-Entropy (CE) method that utilizes a beta distribution to simulate the locations of change points. Furthermore, parallel implementation of the extended CE method results a significant improvement in total processing time, in which the procedures are computationally highly intensive. We consider an artificially generated count data sequence to assess the performance of the propose method. Finally, a real DNA count data set is used to illustrate the usefulness of the proposed methodology.