Title: Statistical assessment of depth normalization methods for microRNA sequencing
Authors: Li-Xuan Qin - Memorial Sloan Kettering Cancer Center (United States) [presenting]
Abstract: Quality data is the foundational cornerstone for reliable scientific findings in evidence-based medical research. It is widely accepted that a crucial step to derive high-quality genomics data is to identify data artifacts caused by systematic differences in the processing of specimens and to remove these artifacts by data normalization. Statistical methods for normalizing sequencing data depth have been recently developed, including re-scaling-based and regression-based methods. Many of these methods rely on the presupposition that variations in the assumed scaling factor or projection of the assumed regression function are solely due to data artifacts and should be removed. MicroRNAs are a unique class of small RNAs closely linked to carcinogenesis. They are low-complexity molecules that tend to be expressed in a tissue-specific manner. As a result, the assumption of depth normalization methods may not hold for microRNA sequencing. We performed a study to assess the performance of existing depth normalization methods on identifying disease relevant microRNAs using both a pair of datasets on the same set of tumor samples and data simulated from this dataset pair under various scenarios of differential expression. We will report our findings from this study.