Bernie J. Daigle

Independent component analysis: Mining microarray data for fundamental human gene expression modules

As public microarray repositories rapidly accumulate gene expression data, these resources contain increasingly valuable information about cellular processes in human biology. This presents a unique opportunity for intelligent data mining methods to extract information about the transcriptional modules underlying these biological processes. Modeling cellular gene expression as a combination of functional modules, we use independent component analysis (ICA) to derive 423 fundamental components of human biology from a 9395-array compendium of heterogeneous expression data. Annotation using the Gene Ontology (GO) suggests that while some of these components represent known biological modules, others may describe biology not well characterized by existing manually-curated ontologies. In order to understand the biological functions represented by these modules, we investigate the mechanism of the preclinical anti-cancer drug parthenolide (PTL) by analyzing the differential expression of our fundamental components. Our method correctly identifies known pathways and predicts that N-glycan biosynthesis and T-cell receptor signaling may contribute to PTL response. The fundamental gene modules we describe have the potential to provide pathway-level insight into new gene expression datasets.

Automated estimation of rare event probabilities in biochemical systems.

In biochemical systems, the occurrence of a rare event can be accompanied by catastrophic consequences. Precise characterization of these events using Monte Carlo simulation methods is often intractable, as the number of realizations needed to witness even a single rare event can be very large. The weighted stochastic simulation algorithm (wSSA) [J. Chem. Phys. 129, 165101 (2008)] and its subsequent extension [J. Chem. Phys. 130, 174103 (2009)] alleviate this difficulty with importance sampling, which effectively biases the system toward the desired rare event. However, extensive computation coupled with substantial insight into a given system is required, as there is currently no automatic approach for choosing wSSA parameters. We present a novel modification of the wSSA--the doubly weighted SSA (dwSSA)--that makes possible a fully automated parameter selection method. Our approach uses the information-theoretic concept of cross entropy to identify parameter values yielding minimum variance rare event probability estimates. We apply the method to four examples: a pure birth process, a birth-death process, an enzymatic futile cycle, and a yeast polarization model. Our results demonstrate that the proposed method (1) enables probability estimation for a class of rare events that cannot be interrogated with the wSSA, and (2) for all examples tested, reduces the number of runs needed to achieve comparable accuracy by multiple orders of magnitude. For a particular rare event in the yeast polarization model, our method transforms a projected simulation time of 600 years to three hours. Furthermore, by incorporating information-theoretic principles, our approach provides a framework for the development of more sophisticated influencing schemes that should further improve estimation accuracy.

Accelerated maximum likelihood parameter estimation for stochastic biochemical systems

BackgroundA prerequisite for the mechanistic simulation of a biochemical system is detailed knowledge of its kinetic parameters. Despite recent experimental advances, the estimation of unknown parameter values from observed data is still a bottleneck for obtaining accurate simulation results. Many methods exist for parameter estimation in deterministic biochemical systems; methods for discrete stochastic systems are less well developed. Given the probabilistic nature of stochastic biochemical models, a natural approach is to choose parameter values that maximize the probability of the observed data with respect to the unknown parameters, a.k.a. the maximum likelihood parameter estimates (MLEs). MLE computation for all but the simplest models requires the simulation of many system trajectories that are consistent with experimental data. For models with unknown parameters, this presents a computational challenge, as the generation of consistent trajectories can be an extremely rare occurrence.ResultsWe have developed Monte Carlo Expectation-Maximization with Modified Cross-Entropy Method (MCEM2): an accelerated method for calculating MLEs that combines advances in rare event simulation with a computationally efficient version of the Monte Carlo expectation-maximization (MCEM) algorithm. Our method requires no prior knowledge regarding parameter values, and it automatically provides a multivariate parameter uncertainty estimate. We applied the method to five stochastic systems of increasing complexity, progressing from an analytically tractable pure-birth model to a computationally demanding model of yeast-polarization. Our results demonstrate that MCEM2 substantially accelerates MLE computation on all tested models when compared to a stand-alone version of MCEM. Additionally, we show how our method identifies parameter values for certain classes of models more accurately than two recently proposed computationally efficient methods.ConclusionsThis work provides a novel, accelerated version of a likelihood-based parameter estimation method that can be readily applied to stochastic biochemical systems. In addition, our results suggest opportunities for added efficiency improvements that will further enhance our ability to mechanistically simulate biological processes.

Mice lacking the β2 adrenergic receptor have a unique genetic profile before and after focal brain ischaemia

The role of the β2AR (β2 adrenergic receptor) after stroke is unclear as pharmacological manipulations of the β2AR have produced contradictory results. We previously showed that mice deficient in the β2AR (β2KO) had smaller infarcts compared with WT (wild-type) mice (FVB) after MCAO (middle cerebral artery occlusion), a model of stroke. To elucidate mechanisms of this neuroprotection, we evaluated changes in gene expression using microarrays comparing differences before and after MCAO, and differences between genotypes. Genes associated with inflammation and cell deaths were enriched after MCAO in both genotypes, and we identified several genes not previously shown to increase following ischaemia (Ccl9, Gem and Prg4). In addition to networks that were similar between genotypes, one network with a central core of GPCR (G-protein-coupled receptor) and including biological functions such as carbohydrate metabolism, small molecule biochemistry and inflammation was identified in FVB mice but not in β2KO mice. Analysis of differences between genotypes revealed 11 genes differentially expressed by genotype both before and after ischaemia. We demonstrate greater Glo1 protein levels and lower Pmaip/Noxa mRNA levels in β2KO mice in both sham and MCAO conditions. As both genes are implicated in NF-κB (nuclear factor κB) signalling, we measured p65 activity and TNFα (tumour necrosis factor α) levels 24 h after MCAO. MCAO-induced p65 activation and post-ischaemic TNFα production were both greater in FVB compared with β2KO mice. These results suggest that loss of β2AR signalling results in a neuroprotective phenotype in part due to decreased NF-κB signalling, decreased inflammation and decreased apoptotic signalling in the brain.

M-BISON: Microarray-based integration of data sources using networks

BackgroundThe accurate detection of differentially expressed (DE) genes has become a central task in microarray analysis. Unfortunately, the noise level and experimental variability of microarrays can be limiting. While a number of existing methods partially overcome these limitations by incorporating biological knowledge in the form of gene groups, these methods sacrifice gene-level resolution. This loss of precision can be inappropriate, especially if the desired output is a ranked list of individual genes. To address this shortcoming, we developed M-BISON (Microarray-Based Integration of data SOurces using Networks), a formal probabilistic model that integrates background biological knowledge with microarray data to predict individual DE genes.ResultsM-BISON improves signal detection on a range of simulated data, particularly when using very noisy microarray data. We also applied the method to the task of predicting heat shock-related differentially expressed genes in S. cerevisiae, using an hsf1 mutant microarray dataset and conserved yeast DNA sequence motifs. Our results demonstrate that M-BISON improves the analysis quality and makes predictions that are easy to interpret in concert with incorporated knowledge. Specifically, M-BISON increases the AUC of DE gene prediction from .541 to .623 when compared to a method using only microarray data, and M-BISON outperforms a related method, GeneRank. Furthermore, by analyzing M-BISON predictions in the context of the background knowledge, we identified YHR124W as a potentially novel player in the yeast heat shock response.ConclusionThis work provides a solid foundation for the principled integration of imperfect biological knowledge with gene expression data and other high-throughput data sources.

Whole-genome DNA methylation status associated with clinical PTSD measures of OIF/OEF veterans

Emerging knowledge suggests that post-traumatic stress disorder (PTSD) pathophysiology is linked to the patients’ epigenetic changes, but comprehensive studies examining genome-wide methylation have not been performed. In this study, we examined genome-wide DNA methylation in peripheral whole blood in combat veterans with and without PTSD to ascertain differentially methylated probes. Discovery was initially made in a training sample comprising 48 male Operation Enduring Freedom (OEF)/Operation Iraqi Freedom (OIF) veterans with PTSD and 51 age/ethnicity/gender-matched combat-exposed PTSD-negative controls. Agilent whole-genome array detected ~5600 differentially methylated CpG islands (CpGI) annotated to ~2800 differently methylated genes (DMGs). The majority (84.5%) of these CpGIs were hypermethylated in the PTSD cases. Functional analysis was performed using the DMGs encoding the promoter-bound CpGIs to identify networks related to PTSD. The identified networks were further validated by an independent test set comprising 31 PTSD+/29 PTSD− veterans. Targeted bisulfite sequencing was also used to confirm the methylation status of 20 DMGs shown to be highly perturbed in the training set. To improve the statistical power and mitigate the assay bias and batch effects, a union set combining both training and test set was assayed using a different platform from Illumina. The pathways curated from this analysis confirmed 65% of the pool of pathways mined from training and test sets. The results highlight the importance of assay methodology and use of independent samples for discovery and validation of differentially methylated genes mined from whole blood. Nonetheless, the current study demonstrates that several important epigenetically altered networks may distinguish combat-exposed veterans with and without PTSD.

A Multivariate Ensemble Approach for Identification of Biomarkers: Application to Breast Cancer

Abstract Advances in high throughput screening experiments have significantly improved our ability to discover and predict biomarkers for complex diseases. Systems biology approaches have played a critical role in realizing these improvements by providing computational tools for modeling such diseases at the network level. Within these tools, statistical scores such as the two sample t-statistic (t-score) are commonly used to rank genes/features for downstream analyses. In this paper, we propose a new alternative to the t-score —the ensemble sensitivity (ES) metric —which is a multivariate strategy to obtain feature rankings. To validate our method, we employ the COre Module Biomarker Identification with Network Exploration (COMBINER) tool on publicly available breast cancer gene expression data sets. Top candidates obtained by both the t-score and ES method serve as an input to COMBINER, which identifies the candidate biomarkers. Our results, as quantified by the COMBINER-generated area under the ROC curve (AUC), suggest that the ES approach improves the average AUC and identifies biomarkers with ~ 93% overlap with known cancer-related genes. In addition, the overlap of genes known to be associated with cancer that are identified using the two methods is small. This suggests that our proposed approach captures signals missed by methods relying on the t-score.

Core module network construction for breast cancer metastasis

For prognostic and diagnostic purposes, it is crucial to be able to separate the group of “driver” genes and their first-degree neighbours, (i.e. “core module”) from the general “disease module”. To facilitate this task, we developed a novel computational framework COMBINER: COre Module Biomarker Identification with Network ExploRation. We applied COMBINER to three benchmark breast cancer datasets for identifying prognostic biomarkers. We generated a list of “driver genes” by finding the common core modules between two sets of COMBINER markers identified with different module inference protocols. Overlaying the markers on the map of “the hallmarks of cancer” and constructing a weighted regulatory network with sensitivity analysis, we validated 29 driver genes. Our results show the COMBINER framework to be a promising approach for identifying and characterizing core modules and driver genes of many complex diseases.