Measuring Reproducibility of High-Throughput Deep-Sequencing Experiments Based on Self-adaptive Mixture Copula

Measurement of the statistical reproducibility between biological experiment replicates is vital first step of the entire series of bioinformatics analysis for mining meaningful biological discovery from mega-data. To distinguish the real biological relevant signals from artificial signals, irreproducible discovery rate (IDR) employing Copula, which can separate dependence structure and marginal distribution from data, has been put forth. However, IDR employed a Gaussian Copula which may cause underestimation of risk and limit the robustness of the method. To address the issue, we propose a Self-adaptive Mixture Copula (SaMiC) to measure the reproducibility of experiment replicates from high-throughput deep-sequencing data. Simple and easy to implement, the proposed SaMiC method can self-adaptively tune its coefficients so that the measurement of reproducibility is more effective for general distributions. Experiments in simulated and real data indicate that compared with IDR, the SaMiC method can better estimate reproducibility between replicate samples.