Standardized median difference for quality control in high-throughput screening

High-throughput screening (HTS) of large-scale ribonucleic acid interference (RNAi) libraries has been widely used in functional genomics in recent years. The Z-factor and strictly standardized mean difference (SSMD) have already become two popular methods for quality control (QC) in HTS experiments. This paper is intended to put forward a robust parameter, standardized median difference (SMD), for group comparison and derive some statistical properties of SMD, such as maximum likelihood estimate (MLE), moment-method (MM) estimate and plug-in estimate of SMD as well as their relative asymptotical distributions. Using the simulations, the plug-in estimate of SMD obtains QC results robust to outliers whereas the Z-factor and SSMD obtain QC results sensitive to outliers. Meanwhile, compared to other QC metrics, the SMD also obtains non-inferior QC results when there are no outliers.