Time Series Analysis of Gene Expression and Location Data.pdfVIP

Time Series Analysis of Gene Expression and Location Data Chen-Hsiang Yeang Tommi Jaakkola Artificial Intelligence Lab, MIT. Artificial Intelligence Lab, MIT. Cambirdge, MA 02139, USA. Cambridge, MA 02139, USA. chyeang@ tommi@ Abstract expression analysis [6, 2, 1] and the combination of such data with location or other data sources such as sequence We develop a method for integrating time series expres- motifs [5, 11, 12, 13]. However, systematic approaches of sion profiles and factor-gene binding data to quantify dy- combining properties of time course profiles with other ge- namic aspects of gene regulation. We estimate latencies nomic data are relatively scarce in previous studies. While, for transcription activation by explaining time correlations for example, [11, 12] certainly fall in this category, their fo- between gene expression profiles through available factor- cus were more on using one data source as a means of val- gene binding information. The resulting aligned expres- idating the other. To fully use time course profiles it seems sion profiles are subsequently clustered and again combined necessary to temporally align individual gene profiles. with binding information to determine groups or subgroups The contribution of this paper is twofold. First, we pro- of co-regulated genes. The predictions derived from this ap- pose an algorithm to