Tackling Computational Challenges in High-Throughput RNA Interference Screening

Abstract

Since the discovery of RNAi decades ago, it has been increasingly used in biomedical and biological research. The success of analyzing single genes using siRNAs has resulted in the large-scale application of RNAi for genome-wide loss-of-function phenotype screening while reducing cost and decreasing time. High-throughput RNAi screening (HTS) has been widely accepted and used in a variety of biomedical and biological research projects as the first step to identifying novel drug targets or pathway components. Huge data sets are being generated, but computational challenges remain in data analysis and hit identification, which have become hurdles in HTS. These must be tackled before we can more accurately and precisely interpret the HTS results, since they are often blurred by spatial noise and off-target effects. In my thesis research, I have been working on statistical modeling of high-throughput RNAi screening results. I developed SbacHTS (spatial background noise correction in high-throughput RNAi screening) to identify and remove spatially-correlated background noise from HTS, which helps enhance statistical detection power in triplicate experiments. On top of that, I also created a novel algorithm, DeciRNAi (deconvolution analysis high-throughput RNAi screening results), to quantify the strength and direction of siRNA-mimic-miRNA off-target effects in HTS projects. As a special case, image-based high-content HTS requires management of high-dimensional data analysis and visualization. I built a new R package “iScreen” (image-based high-throughput RNAi screening analysis tools) to deal with such problems.