Inferring Time-Lagged Association Rules from Microarray Time-Series Data

GALLO, CRISTIAN A.; CARBALLIDO, JESSICA A.; PONZONI, IGNACIO

Santiago

Conferencia; 1st International Conference in Bioinformatics; 2010

Sociedad Iberoamericana de Bioinformática (SoiBio)

Gene regulatory networks have an important role in every process of life, including cell differentiation, metabolism, the cell cycle and signal transduction. The amount of large scale and genome wide time-series data is becoming increasingly available, which provides the opportunity to reverse engineer the time-delayed gene regulatory networks that govern the majority of these molecular processes. In particular, this paper aims to reconstruct regulatory networks from microarray time-series data by using delayed correlations between genes. Thereby, a new model-free computational toolbox termed GRNCOP2 (Gene Regulatory Network inference by Combinatorial OPtimization v. 2) was developed in order to address the underlying association between genes that can span any unit(s) of time intervals. This toolbox overcomes the limitations of the original version1 by means of several improvements. The proposed method was applied to a time-series data of 20 yeast genes and the results were compared against several approaches available in the literature1, 2, 3. In order to assess and contrast the performance of the methods, the proportion of associations (i.e., rules without repetition) that share any GO biological process terms of a Gene Ontology annotation downloaded from the Saccharomyces cerevisiae Genome Database4 was measured. Figure 1 shows the scores achieved for the algorithms with an accuracy of 0.75 for simultaneous and time-delayed rules.  As it can be observed, GRNCOP2 outperforms all the referential methods in terms of the proportion of rules that validate with the GO annotations. This demonstrates that the improvements incorporated in this new version of the algorithm provide a usable and powerful model-free approach to dissecting high-order dynamic trends of gene-gene interactions, carefully validated with publicly available cell cycling datasets through the well recognized S. cerevisiae Genome Database.