We describe a Bayesian network approach to infer genetic regulatory interactions from microarray gene expression data. This problem was introduced in Chapter 7 and an alternative Bayesian network approach was presented in Chapter 8. Our approach is based on a linear dynamical system, which renders the inference problem tractable: the E-step of the EM algorithm draws on the well-established Kalman smoothing algorithm. While the intrinsic linearity constraint makes our approach less suitable for modeling non-linear genetic interactions than the approach of Chapter 8, it has two important advantages over the method of Chapter 8. First, our approach works with continuous expression levels, which avoids the information loss inherent in a discretization of these signals. Second, we include hidden states to allow for the effects that cannot be measured in a microarray experiment, for example: the effects of genes that have not been included on the microarray, levels of regulatory proteins, and the effects of mRNA and protein degradation.
CITATION STYLE
Probabilistic Modeling in Bioinformatics and Medical Informatics. (2005). Probabilistic Modeling in Bioinformatics and Medical Informatics. Springer London. https://doi.org/10.1007/b138794
Mendeley helps you to discover research relevant for your work.