B1784
Title: RUV statistical methods based on generalised linear models for omics data
Authors: Alysha De Livera - La Trobe University (Australia) [presenting]
Terry Speed - Walter and Eliza Hall Institute of Medical Research (Australia)
Agus Salim - The University of Melbourne (Australia)
Abstract: Unwanted variations in omics data, not only inevitably arise from various technical sources, such as the use of multiple analytical platforms, batches, laboratories, long-run of samples and temperature changes within instruments, but also from unwanted biological variations, such as different cell sizes which are often unmeasurable. Failure to carry out a suitable approach to removing unwanted variation (RUV) in the statistical analysis of omics data, leads to increases in Type I and Type II errors, spurious correlation, as well as artificial clustering and poor classification of the biological samples. Over the last decade, RUV statistical methods have been established as popular, widely-used methods in multiple omics fields for removing unwanted variation. We will go through the latest developments in the RUV methods based on generalised linear models, demonstrate their applications to RNA-sequencing and single-cell data, and compare their performance with the existing methods as well as RUV counterparts which are based on linear models.
