Over the last few years, many microarray-based gene expression studies involving a large number of samples have been carried out, with the hope of understanding, predicting or discovering factors of interest such as prognosis or the subtypes of a cancer. The same applies to proteomic and metabolomic data, and to other kinds of data. Such large studies are often carried out over several years, and may involve several hospitals or research centers. Unwanted variation (UV) can arise from technical aspects such as batches, different platforms or laboratories, or from biological signals such as heterogeneity in ages or different ethnic groups which are unrelated to the factor of interest in the study. This can easily lead to poor results. Recently, we proposed a general framework to remove UV (called RUV) in microarray data using control genes. It showed good behavior for differential expression analysis (i.e., with a known factor of interest) when applied to several datasets, in particular better performance than state of the art methods such as Combat or SVA. This suggests that controls can indeed be used to estimate and efficiently remove sources of unwanted variation. The methods are illustrated on a variety of kinds of omic datasets.
About the 2014 AMSI-SSAI Lecture Tour:
Between the months of August and November this year, the 2013 Prime Minister’s Science Prize winner and one of Australia’s most eminent statisticians, Professor Terry Speed, will be touring the country as the 2014 AMSI-SSAI Lecturer. This AGR Seminar is part of the Lecture Tour.
This annual event gives the research community and the general public an opportunity to hear top academics in the fields of both pure and applied mathematics speak about their research.
For more information about the Lecture Tour schedule, please click here.