Prioritization of cell cycle regulated genes

Abstract

The cell cycle is an important biological process in which a set of events occurs in a sequential manner progressing to the cell division. Cell cycle is regulated by periodic fluctuations in the expression levels of several genes referred to as cell cycle regulated genes. In this study, we apply machine learning techniques to prioritize a list of candidate genes with respect to being involved in the cell cycle regulation process. We focus on the data obtained from different expression experiments on which partial least squares regression (PLS) models have been previously developed to identify genes with cell cycle dependent expression profiles. The different expression experiments used different synchronization methods to halt the cell cultures, so that each experiment started to measure gene expression values at different cell cycle phases after synchronization. We are mainly interested in genes having cyclic expression profile which is consistent with respect to cell cycle phases within all experiments. Our goal is therefore to develop a method that can identify genes that have consistent cyclic expression profiles across multiple synchronization experiments.We solve the cell cycle related gene prioritization problem through a novelty detection approach using one-class support vector machine. The candidate genes are ranked according to their similarity to the genes with known cell cycle function. After checking the function of the top ranked genes, it is found that most of them are involved in biological processes related to the cell cycle, which is a good indication that our approach is able to prioritize genes with cell cycle function.Keywords: Cell cycle, Partial least squares (PLS), Gene prioritization, one-class support vector machine.