A modified correlation coefficient based similarity measure for clustering time-course gene expression data

On the evaluation of clustering results: measures, ensembles, and gene expression data analysis

“…In such experiments, usually referred to as time-course, each sample provides a snapshot of the cell's state in a different instant of time. Because of the time component involved, each gene under analysis can be regarded as a time-series, although traditional time-series techniques usually cannot by employed due to their short length (Son and Baek, 2008 Figure 3.4: Depiction of a gene expression data matrix. Each biological sample from the matrix comes from a different experiment.…”

“…GENE ONTOLOGY literature. Additionally, some distances have been specifically introduced aiming the clustering of gene time-series, e.g., (Balasubramaniyan et al, 2005;Heyer et al, 1999;Möller-Levet et al, 2005;Son and Baek, 2008), taking into account its temporal characteristics. In this context, in Chapter 6 we introduce a methodology for the evaluation of distance measures based on the Gene Ontology (Ashburner et al, 2000).…”

“…Although this somehow limits the number of choices from the whole universe of distance measures, there is still a considerable variety of measures capable of identifying trend similarity available in the general clustering literature. Additionally, some distances have been specifically introduced aiming the clustering of gene time-series, e.g., Heyer et al (1999), Balasubramaniyan et al (2005), Möller-Levet et al (2005), and Son and Baek (2008), taking into account the temporal characteristics involved in the experiments. It is worth noticing that despite the variety of distance measures available for the clustering of gene expression data, only a handful of works tried to provide guidelines concerning their choice, as we review in the following.…”

“…Based on the presumption that correlations may not capture all information contained in gene time-series, Son and Baek (2008) introduced two dissimilarities that combine different types of information with correlation values. The first information taken into account concerns the agreement between the m − 1 slopes that compose the two gene series under comparison, as given by Equation (6.15).…”

“…In Equations (6.19) and (6.20), the weight terms ω 1 , ω 2 , and ω 3 must satisfy the condition Son and Baek (2008) proposed a way to estimate their values, but their approach is highly computationally demanding. To compare the two measures we set weights in the following form: ω 1 = 0.5, ω 2 = 0.25, and ω 3 = 0.25, which are values employed by the authors in (Son and Baek, 2008).…”