Transcriptome Data Analysis for Cell Culture Processes

Abstract

In the past decade, DNA microarrays have fundamentally changed the way we study complex biological systems. By measuring the expression levels of thousands of transcripts, the paradigm of studying organisms has shifted from focusing on the local phenomena of a few genes to surveying the whole genome. DNA microarrays are used in a variety of ways, from simple comparisons between two samples to more intricate time-series studies. With the large number of genes being studied, the dimensionality of the problem is inevitably high. The analysis of microarray data thus requires specific approaches. In the case of time-series microarray studies, data analysis is further complicated by the correlation between successive time points in a series.In this review, we survey the methodologies used in the analysis of static and time-series microarray data, covering data pre-processing, identification of differentially expressed genes, profile pattern recognition, pathway analysis, and network reconstruction. When available, examples of their use in mammalian cell cultures are presented.

Analysis of time-course gene expression profiles to study regulation of cell growth in fed-batch bioreactors

“…Tools such as genomic microarrays can help develop a better understanding of how changes in the cell culture environment relate to the biological state of cells [8,9]. Since gene expression is only one facet of the information contained within cells, this analysis alone likely does not provide a full story but instead establishes a good starting point to develop testable hypothesis and begin to comprehend the complex underlying mechanisms that drive cell growth in fed-batch bioreactors.…”

“…This study represents one of just a few that have utilized microarray data to draw connections between cell biology and time-course performance data from bioreactors [9]. In this study, we have addressed important challenges associated with microarray data including handling batch-to-batch variability, selecting for a few genes that are relevant to the phenotype of interest and building biological hypothesis and understanding based on the knowledge of these relevant genes and their correlation to the phenotype of interest.…”

“…The mature field of microarray analysis provides one such avenue that can be used to easily generate time-course profiles for mRNA expression in fed-batch bioreactors [8][9][10]. However, there are some technical hurdles that warrant the use of specialized statistical methods before relevant information can be extracted from microarray data.…”

“…Also, gene expression profiles are not independent and are usually covariant with expression of other related genes, which leads to the other issue of multicollinearity [13]. One important strategy to address both the high-dimensionality and multicollinearity problems is to use dimension reduction methods, such as principal component analysis (PCA) and partial least squares (PLS) [9,14]. These approaches reduce dimensionality by replacing the large number of predictors (genes) with just a few significant principal components that capture the major trends in the data.…”

“…However, these approaches do not lead to a reduction in the number of predictors (genes) in the dataset. Since microarrays yield information on several thousand expressed genes in the cells, it is especially difficult to identify genes of interest in context of the phenotype under consideration [9,15]. In order to address this, methods that allow for selection of relevant predictors (genes) within the dataset have been proposed.…”