Unraveling complex temporal associations in cellular systems across multiple time-series microarray datasets

Abstract

Unraveling the temporal complexity of cellular systems is a challenging task, as the subtle coordination of molecular activities cannot be adequately captured by simple mathematical concepts such as correlation. This paper addresses the challenge with a data-mining approach. We introduce the novel concept of a “frequent temporal association pattern” (FTAP): a set of genes simultaneously exhibit complex temporal expression patterns recurrently across multiple microarray datasets. Such temporal signals are hard to identify in individual microarray datasets, but become significant by their frequent occurrences across multiple datasets. We designed an efficient two-stage algorithm to identify FTAPs. First, for each gene we identify expression trends that occur frequently across multiple datasets. Second, we look for a set of genes that simultaneously exhibit their respective trends recurrently in multiple datasets. We applied this algorithm to 18 yeast time-series microarray datasets. The majority of FTAPs identified by the algorithm are associated with specific biological functions. Moreover, a significant number of patterns include genes that are functionally related but do not exhibit co-expression; such gene groups cannot be captured by clustering algorithms. Our approach offers advantages: (1) it can identify complex associations of temporal trends in gene expression, an important step towards understanding the complex mechanisms governing cellular systems; (2) it is capable of integrating time-series data with different time scales and intervals; and (3) it yields results that are robust against outliers.