Inferring Genetic Interactions via a Pattern Recognition Approach

Abstract

A brief introduction to the area of predicting genetic networks, including why it is one of the frontier topics in computational biology, will be given first. Next, I will focus on one of the algorithms developed by my team PARE (Chuang et al., 2008), and its web-computing service (WebPARE, Chuang et al., 2009). For any time-course gene expression data in which the gene interactions and the associated paired patterns are dependent, a pattern recognition (PARE) approach can infer genetic interactions, a challenging task due to the small number of time points and large number of genes. PARE uses a non-linear score to identify subclasses of gene pairs with different time lags. In each subclass, PARE extracts non-linear characteristics of paired gene-expression curves and learns the parameters in the decision score using expression data and known interactions, from biological experiments or published literature. PARE, a time-lagged correlation approach and the latest advance in graphical Gaussian models were applied to predict 112 (132) pairs of genetic (transcriptional regulatory) interactions. Checked against qRT-PCR results (published literature), their true positive rates are 73% (77%), 46% (51%), and 52% (59%), respectively; the false positive rates for inferring the yeast genome are bounded by 23% (24%), respectively. Several predicted TC/TD interactions are shown to coincide with existing pathways, which shows PARE may be applied to predict biochemical pathways. Further, PARE was applied to infer transcriptional interactions involved in human obesity. Finally, WebPARE (http://www.stat.sinica.edu.tw/WebPARE/) will also be demonstrated if time permits.