Bayesian State Space Models for Inferring and Predicting Temporal Gene Expression Profiles

Abstract

Prediction of gene dynamic behavior is a challenging and important problem in genomic research while estimating the temporal correlations and non-stationarity are the keys in this process. Unfortunately, most existing techniques used for the inclusion of the temporal correlations treat the time course as evenly distributed time intervals and use stationary models with time-invariant settings. This is an assumption that is often violated in microarray time course data since the time course expression data are at unequal time points, where the difference in sampling times varies from minutes to days. Furthermore, the unevenly spaced short time courses with sudden changes make the prediction of genetic dynamics difficult. In this paper, we develop two types of Bayesian state space models to tackle this challenge for inferring and predicting the gene expression profiles associated with diseases. In the univariate time-varying Bayesian state space models we treat both the stochastic transition matrix and the observation matrix time-variant with linear setting and point out that this can easily be extended to nonlinear setting. In the multivariate Bayesian state space model we include temporal correlation structures in the covariance matrix estimations. In both models, the unevenly spaced short time courses with unseen time points are treated as hidden state variables. Bayesian approaches with various prior and hyper-prior models with MCMC algorithms are used to estimate the model parameters and hidden variables. We apply our models to multiple tissue polygenetic affymetrix data sets. Results show that the predictions of the genomic dynamic behavior can be well captured by the proposed models.

Transcriptional profiling of VEGF-A and VEGF-C target genes in lymphatic endothelium reveals endothelial-specific molecule-1 as a novel mediator of lymphangiogenesis

“…To analyze gene expression of triplicate samples at 5 different time points, we applied the multivariate EB approach to inference. 44,45 This is a model-based strategy for introducing moderation into the analysis that has been reported to generate the least number of false positives and false negatives. 26 Using STEM analysis, 29 we identified distinct temporal clusters of genes modulated by VEGF-A and VEGF-C.…”

“…25,[40][41][42][43] We investigated the genes modulated by VEGF-A and/or VEGF-C by applying multivariate EB statistics, which ranks genes on the basis of their sequential expression over time and the reproducibility at each time point. [44][45][46][47] We next performed STEM analysis 29 to determine which significantly modulated genes cluster together based on their temporal regulation pattern (Figure 1). For the VEGF-A-treated LECs, we identified 97 genes clustering into the early-response genes group (peak at time point 1 hour), 51 genes into the transiently induced genes (peak between 4 and 8 hours), 87 genes into the progressively induced genes group (progressive increase of expression over time), and 59 genes into the down-regulated genes group (progressive decrease over time).…”

Big Data Science and its Applications in Healthcare and Medical Research: Challenges and Opportunities

“…For instance, in the genomic domain, for Affymetrix time course data obtained from Affymetrix GeneChips, one may use Affymetrix software (MAS 5.0) and probe set algorithms of MAS5 for background subtraction, signal intensity normalization between arrays, and non-specific hybridization correction etc [75][76][77][78][79]. To do so, high level performance hardware and software (e.g., programming languages and algorithms for visualizations) that conduct parallel and distributed and cloud computing to manage, retrieve, reformat and analyze the data from various resources including the genomic laboratory and hospital patient information systems needs be considered (Table 1) [58,59,[80][81][82][83][84].…”

“…Moreover, post clustering and path analysis is able to not only identify the genes that are over expressed, under-or not expressed, but to isolate trajectories of genes whose regulations appear to be interdependent, inferring the possible inter-gene-dependence pathway and network showing early, intermediate, and late gene clusters to better understand the treatment effect. In short, the combinations of these various approaches would provide us more comprehensive picture of the solutions and reliable results that illustrates the values and roles of the advanced computational tools transforming thousands of Big Data points into quantitative statistical evidence for diagnostics, therapeutics, and new insights into disease, population health, and treatment [75][76][77][78][79][85][86][87]. Health/nursing and medical researchers could employ these advanced analytical tools in Big genome research for either disease specific (e.g., neurology conditions, cancer, cardiovascular diseases) or domain specific such as pain, fatigue, physical functioning or multiple chronic health conditions.…”

Bayesian state space models for dynamic genetic network construction across multiple tissues

“…In our earlier studies, we have developed and applied univariate Bayesian state space model for modeling the temporal gene expression data, which have advantages of uniformly integrating the stochastic paradigm with the observation data to capture the change of gene states (Liang and Kelemen, 2007). Stochastic paradigm treats the observation process of gene expression change as a stochastic process to describe the evolution of a probability genetic system in time.…”