253 – Contributed Oral Poster Presentations: Biometrics
Statistical Approaches for Off-Target Effects Identification and Corrections in High-Throughput RNAi Screenings
Michael White
U.S. Census Bureau
Andy Xiao
The University of Texas Southwestern Medical Center
Yang Xie
The University of Texas Southwestern Medical Center
Rui Zhong
The University of Texas Southwestern Medical Center
High-throughput RNA-interference (RNAi) screening has been widely used in biological research, like discovery of unknown molecular machinery and identification of novel drug-targetable genes. However, off-target effects make interpretation blurry. Here we develop a novel computational approach to identify off-targeting siRNA oligos based on miRNA-mimic mechanism. Genome-wide siRNA oligos are classified into different seed families based on their seed sequences. We used KS (Kolmogorov-Smirnov) test to determined enrichment for each seed family. We modeled each Z score as a linear combination of seed families' off-target effects and on-target effects, and then estimated the off-target using penalized regression with LASSO (least absolute shrinkage and selection operator) penalty term. Using the modeling approach, we could adjust off-target effects from the original Z scores and our results showed that corrected Z scores improved accuracy in hits selection. In a real data application to identify selective autophagy factors, our method led to hits with higher confirmation rate in the secondary confirmation screening.