Pathway Membership Research Articles

Hierarchical Feature Selection Incorporating Known and Novel Biological Information: Identifying Genomic Features Related to Prostate Cancer Recurrence

ABSTRACTOur work is motivated by a prostate cancer study aimed at identifying mRNA and miRNA biomarkers that are predictive of cancer recurrence after prostatectomy. It has been shown in the literature that incorporating known biological information on pathway memberships and interactions among biomarkers improves feature selection of high-dimensional biomarkers in relation to disease risk. Biological information is often represented by graphs or networks, in which biomarkers are represented by nodes and interactions among them are represented by edges; however, biological information is often not fully known. For example, the role of microRNAs (miRNAs) in regulating gene expression is not fully understood and the miRNA regulatory network is not fully established, in which case new strategies are needed for feature selection. To this end, we treat unknown biological information as missing data (i.e., missing edges in graphs), different from commonly encountered missing data problems where variable values are missing. We propose a new concept of imputing unknown biological information based on observed data and define the imputed information as the novel biological information. In addition, we propose a hierarchical group penalty to encourage sparsity and feature selection at both the pathway level and the within-pathway level, which, combined with the imputation step, allows for incorporation of known and novel biological information. While it is applicable to general regression settings, we develop and investigate the proposed approach in the context of semiparametric accelerated failure time models motivated by our data example. Data application and simulation studies show that incorporation of novel biological information improves performance in risk prediction and feature selection and the proposed penalty outperforms the extensions of several existing penalties. Supplementary materials for this article are available online.

Network-based pathway enrichment analysis with incomplete network information.

Pathway enrichment analysis has become a key tool for biomedical researchers to gain insight into the underlying biology of differentially expressed genes, proteins and metabolites. It reduces complexity and provides a system-level view of changes in cellular activity in response to treatments and/or in disease states. Methods that use existing pathway network information have been shown to outperform simpler methods that only take into account pathway membership. However, despite significant progress in understanding the association amongst members of biological pathways, and expansion of data bases containing information about interactions of biomolecules, the existing network information may be incomplete or inaccurate and is not cell-type or disease condition-specific. We propose a constrained network estimation framework that combines network estimation based on cell- and condition-specific high-dimensional Omics data with interaction information from existing data bases. The resulting pathway topology information is subsequently used to provide a framework for simultaneous testing of differences in expression levels of pathway members, as well as their interactions. We study the asymptotic properties of the proposed network estimator and the test for pathway enrichment, and investigate its small sample performance in simulated and real data settings. The proposed method has been implemented in the R-package netgsa available on CRAN. jinma@upenn.eduSupplementary information: Supplementary data are available at Bioinformatics online.

Transcriptomic profile of leaf tissue from the leguminous tree, Millettia pinnata

Millettia pinnata (formerly Pongamia pinnata) is a fast-growing leguminous tree indigenous to the Indian subcontinent, Southeast Asia, and Australia. This species has been introduced to subtropical and arid regions of Africa, India, the Philippines, Malaysia, Australia, and the USA for commercial growth. Exhibiting saline and drought tolerance, as well as nitrogen-fixing properties, M. pinnata has been used extensively for traditional medicine and agriculture and, more recently, for the production of a biofuel feedstock. The large size, high oil content, and fatty acid profile of the seeds are well suited for biofuel production. In this study, we characterized the leaf transcriptome that was assembled de novo from 72 seedlings pooled into eight libraries. Deep paired-end short-read sequencing was performed on individual libraries using the Illumina HiSeq 2000 platform. The Trinity-assembled transcriptome of 25,146 unique genes was annotated with a combination of open-source tools. Functional annotation was facilitated through sequence homology searches, Gene Ontology term assignment, and protein domain identification. A total of 11,873 genes were classified as full-length, and 22,603 sequences were functionally annotated. Predominate Gene Ontology biological process categories included phosphorylation, metabolic processes, and oxidation-reduction processes. Orthologous gene family analysis identified 19,640 families among the 11 sequenced plant species compared. A total of 4280 were conserved across all species, and 103 were unique to the M. pinnata leaf transcriptome. The unique M. pinnata gene families included transcripts with an array of functions including ubiquitin-like modifier proteins and BED zinc finger proteins with membership in pathways related to salt tolerance and disease resistance.

Combinatorial Cis-regulation in Saccharomyces Species.

Transcriptional control of gene expression requires interactions between the cis-regulatory elements (CREs) controlling gene promoters. We developed a sensitive computational method to identify CRE combinations with conserved spacing that does not require genome alignments. When applied to seven sensu stricto and sensu lato Saccharomyces species, 80% of the predicted interactions displayed some evidence of combinatorial transcriptional behavior in several existing datasets including: (1) chromatin immunoprecipitation data for colocalization of transcription factors, (2) gene expression data for coexpression of predicted regulatory targets, and (3) gene ontology databases for common pathway membership of predicted regulatory targets. We tested several predicted CRE interactions with chromatin immunoprecipitation experiments in a wild-type strain and strains in which a predicted cofactor was deleted. Our experiments confirmed that transcription factor (TF) occupancy at the promoters of the CRE combination target genes depends on the predicted cofactor while occupancy of other promoters is independent of the predicted cofactor. Our method has the additional advantage of identifying regulatory differences between species. By analyzing the S. cerevisiae and S. bayanus genomes, we identified differences in combinatorial cis-regulation between the species and showed that the predicted changes in gene regulation explain several of the species-specific differences seen in gene expression datasets. In some instances, the same CRE combinations appear to regulate genes involved in distinct biological processes in the two different species. The results of this research demonstrate that (1) combinatorial cis-regulation can be inferred by multi-genome analysis and (2) combinatorial cis-regulation can explain differences in gene expression between species.

Abstract C151: Systematic discovery of cancer dependencies through deep coverage pshRNA screens

Abstract Large scale cancer genome sequencing efforts, such as TCGA, revealed the landscape of genomic alterations across many tumor types, but the functional relevance of these alterations and genetic interdependencies cannot be fully assessed from these datasets. Thus, functional genomic screens, such as pooled shRNA 9pshRNA) screens, hold great promise to systematically identify cancer dependencies. However, RNAi discovery screens have been plagued by off target effects causing false positive results and possibly masking on-target effects. In order to address the previous problems with RNAi screens for cancer target discovery, we conducted a large scale pshRNA screening campaign targeting 7500 genes at a depth of 20 shRNAs per gene, across more than 300 cell lines of the cancer cell line encyclopedia (CCLE). The substantial shRNA and cell line depth in this screen significantly increased the robustness of results relative to previous published screens, and allowed the more robust identification of cancer dependencies. Moreover, integration of those growth phenotypes with known features of the cancer cell line encyclopedia (CCLE) enabled the identification of biomarkers (genetic, epigenetic, or proteomic) that correlate with sensitivity, and thus permitted discovery of synthetic lethal relationships. The increased depth provides a robust data set that can be used to find and validate new drug targets as well as infer pathway membership of novel genes by simple phenotypic similarity. The findings of this screen will be presented here. Citation Format: Michael Schlabach, Eric Billy, Konstantinos Mavrakis, Greg Hoffman, Tobias Schmelzle, Francesco Hofmann, Nicholas Keen, Frank Stegmeier, William Sellers. Systematic discovery of cancer dependencies through deep coverage pshRNA screens. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr C151.

Gene expression profiling of ovarian carcinomas and prognostic analysis of outcome.

BackgroundOvarian cancer (OCA), the fifth leading deaths cancer to women, is famous for its low survival rate in epithelial ovarian cancer cases, which is very complicated and hard to be diagnosed from asymptomatic nature in the early stage. Thus, it is urgent to develop an effective genetic prognostic strategy.MethodsCurrent study using the Database for Annotation, Visualization and Integrated Discovery tool for the generation and analysis of quantitative gene expression profiles; all the annotated gene and biochemical pathway membership realized according to shared categorical data from Pathway and Kyoto Encyclopedia of Genes and Genomes; correlation networks based on current gene screening actualize by Weighted correlation network analysis to identify therapeutic targets gene and candidate bio-markers.Results3095 differentially expressed genes were collected from genome expression profiles of OCA patients (n = 53, 35 advanced, 8 early and 10 normal). By pathway enrichment, most genes showed contribution to cell cycle and chromosome maintenance.1073 differentially expression genes involved in the 4 dominant network modules are further generated for prognostic pattern establish, we divided a dataset with random OCA cases (n = 80) into 3 groups efficiently (p = 0.0323, 95 % CIs in Kaplan-Meier). Finally, 6 prognosis related genes were selected out by COX regression analysis, TFCP2L1 related to cancer-stem cell, probably contributes to chemotherapy efficiency.ConclusionsOur study presents an integrated original model of the differentially expression genes related to ovarian cancer progressing, providing the identification of genes relevant for its pathological physiology which can potentially be new clinical markers.Electronic supplementary materialThe online version of this article (doi:10.1186/s13048-015-0176-9) contains supplementary material, which is available to authorized users.

Pinpointing disease genes through phenomic and genomic data fusion.

BackgroundPinpointing genes involved in inherited human diseases remains a great challenge in the post-genomics era. Although approaches have been proposed either based on the guilt-by-association principle or making use of disease phenotype similarities, the low coverage of both diseases and genes in existing methods has been preventing the scan of causative genes for a significant proportion of diseases at the whole-genome level.ResultsTo overcome this limitation, we proposed a rigorous statistical method called pgFusion to prioritize candidate genes by integrating one type of disease phenotype similarity derived from the Unified Medical Language System (UMLS) and seven types of gene functional similarities calculated from gene expression, gene ontology, pathway membership, protein sequence, protein domain, protein-protein interaction and regulation pattern, respectively. Our method covered a total of 7,719 diseases and 20,327 genes, achieving the highest coverage thus far for both diseases and genes. We performed leave-one-out cross-validation experiments to demonstrate the superior performance of our method and applied it to a real exome sequencing dataset of epileptic encephalopathies, showing the capability of this approach in finding causative genes for complex diseases. We further provided the standalone software and online services of pgFusion at http://bioinfo.au.tsinghua.edu.cn/jianglab/pgfusion.ConclusionspgFusion not only provided an effective way for prioritizing candidate genes, but also demonstrated feasible solutions to two fundamental questions in the analysis of big genomic data: the comparability of heterogeneous data and the integration of multiple types of data. Applications of this method in exome or whole genome sequencing studies would accelerate the finding of causative genes for human diseases. Other research fields in genomics could also benefit from the incorporation of our data fusion methodology.

Abstract 4177: Identification of pathways relevant for metastatic site prediction in prostate cancer

Abstract Background and Significance: We address the problem of metastatic site prediction in prostate adenocarcinoma (PRAD), with a specific focus on identifying molecular pathways that are activated in association with the homing to a particular metastatic site. The approach can reveal the molecular mechanisms in metastatic cancer while also providing clues about potential drug targets. Further experimental validation of our findings may lead to the discovery of novel therapies for patients who are in the advanced stages of disease. Methods: We downloaded four PRAD datasets that contained met-site information from the Gene Expression Omnibus (GEO) and trained multi-class predictors on this set. The predictors were then evaluated on patient samples collected as part of the Stand Up To Cancer (SU2C) initiative. Standard normalization techniques were used to remove batch effects associated with non-biological factors such as the institution from which the materials were collected and/or assays conducted. We focused our attention on linear models due to their straightforward interpretation: higher weights indicate stronger association of the corresponding genomic features with a particular metastatic site. To identify pathways implicated by the relevant genomic features, we employed model regularization via group LASSO. This technique groups genes according to their pathway membership using the PathwayCommons database. The regularizer (penalty trading accurate classification with model complexity) sets the weights of an entire group to zero if those groups were uninformative for met-site prediction and non-zero otherwise. Results: We trained a multi-class linear predictor to recognize lymphatic node, liver and bone metastatic sites from gene expression data. The resulting model gave rise to two linear signatures: one that distinguished liver mets from the rest, and another that distinguished lymph node mets from the rest. The signatures were enriched for pathways commonly associated with liver development and liver progenitor cells, as well as pathways involved in integrin interactions on the cell surface. Based on the latter, we hypothesize that the up-regulation of particular integrin-signaling pathways may be responsible for driving the tendency of metastatic PRAD cells to prefer one site over another. We are currently in the process of investigating whether there is further evidence of this hypothesis in the SU2C data, as well as comparing group LASSO to other regularization techniques that also incorporate prior pathway information. Conclusion: We used linear methods to identify several pathways that may be responsible for localization of metastatic prostate adenocarcinoma cells to specific tissues. Our empirical results provide evidence that integrin-signaling may play a key role in this process. We are working on robustness evaluation of these findings, as well as experimental validation with our SU2C collaborators. Citation Format: Adrian Bivol, Kiley Graim, Evan Paull, Dan Carlin, Robert Baertsch, Artem Sokolov, Josh Stuart. Identification of pathways relevant for metastatic site prediction in prostate cancer. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4177. doi:10.1158/1538-7445.AM2014-4177

Exploring the complexity of pathway–drug relationships using latent Dirichlet allocation

Analysis of cellular responses to diverse stimuli enables the exploration in the complexity of functional genomics. Typically, high-throughput microarray data allow us to identify genes that are differentially expressed under a phenomenon of interest. To extract the meanings from the long list of those differentially expressed genes, we present a new method “pathway-based LDA” to determine pathways/gene sets that are perturbed after exposure to different chemicals. In this study, a pathway is defined as a group of functionally related genes. Specifically, we have implemented a probabilistic Latent Dirichlet Allocation (LDA) model to learn drug–pathway–gene relations by taking known gene–pathway memberships as prior knowledge. We applied the pathway-based LDA model and 236 known pathways in order to determine pathway responsiveness to gene expression data of 1169 drugs. Our method yielded a better predictive performance on pathway responsiveness to drug treatments than the existing methods. Moreover, the pathway-based LDA also revealed genes contributing the most in each pre-defined pathway through a probabilistic distribution of genes. In achieving that, our method could provide a useful estimator of the pathway complexity of a genome.

Predicting Protein Relationships to Human Pathways through a Relational Learning Approach Based on Simple Sequence Features.

Biological pathways are important elements of systems biology and in the past decade, an increasing number of pathway databases have been set up to document the growing understanding of complex cellular processes. Although more genome-sequence data are becoming available, a large fraction of it remains functionally uncharacterized. Thus, it is important to be able to predict the mapping of poorly annotated proteins to original pathway models. We have developed a Relational Learning-based Extension (RLE) system to investigate pathway membership through a function prediction approach that mainly relies on combinations of simple properties attributed to each protein. RLE searches for proteins with molecular similarities to specific pathway components. Using RLE, we associated 383 uncharacterized proteins to 28 pre-defined human Reactome pathways, demonstrating relative confidence after proper evaluation. Indeed, in specific cases manual inspection of the database annotations and the related literature supported the proposed classifications. Examples of possible additional components of the Electron transport system, Telomere maintenance and Integrin cell surface interactions pathways are discussed in detail. All the human predicted proteins in the 2009 and 2012 releases 30 and 40 of Reactome are available at http://rle.bioinfo.cnio.es.

Pathway-based Bayesian inference of drug–disease interactions

Drug treatments often perturb the activities of certain pathways, sets of functionally related genes. Examining pathways/gene sets that are responsive to drug treatments instead of a simple list of regulated genes can advance our understanding about such cellular processes after perturbations. In general, pathways do not work in isolation and their connections can cause secondary effects. To address this, we present a new method to better identify pathway responsiveness to drug treatments and simultaneously to determine between-pathway interactions. Firstly, we developed a Bayesian matrix factorisation of gene expression data together with known gene-pathway memberships to identify pathways perturbed by drugs. Secondly, in order to determine the interactions between pathways, we implemented a Gaussian Markov Random Field (GMRF) under the matrix factorization framework. Assuming a Gaussian distribution of pathway responsiveness, we calculated the correlations between pathways. We applied the combination of the Bayesian factor model and the GMRF to analyse gene expression data of 1169 drugs with 236 known pathways, 66 of which were disease-related pathways. Our model yielded a significantly higher average precision than the existing methods for identifying pathway responsiveness to drugs that affected multiple pathways. This implies the advantage of the between-pathway interactions and confirms our assumption that pathways are not independent, an aspect that has been commonly overlooked in the existing methods. Additionally, we demonstrate four case studies illustrating that the between-pathway network enhances the performance of pathway identification and provides insights into disease comorbidity, drug repositioning, and tissue-specific comparative analysis of drug treatments.

Integrating human omics data to prioritize candidate genes

BackgroundThe identification of genes involved in human complex diseases remains a great challenge in computational systems biology. Although methods have been developed to use disease phenotypic similarities with a protein-protein interaction network for the prioritization of candidate genes, other valuable omics data sources have been largely overlooked in these methods.MethodsWith this understanding, we proposed a method called BRIDGE to prioritize candidate genes by integrating disease phenotypic similarities with such omics data as protein-protein interactions, gene sequence similarities, gene expression patterns, gene ontology annotations, and gene pathway memberships. BRIDGE utilizes a multiple regression model with lasso penalty to automatically weight different data sources and is capable of discovering genes associated with diseases whose genetic bases are completely unknown.ResultsWe conducted large-scale cross-validation experiments and demonstrated that more than 60% known disease genes can be ranked top one by BRIDGE in simulated linkage intervals, suggesting the superior performance of this method. We further performed two comprehensive case studies by applying BRIDGE to predict novel genes and transcriptional networks involved in obesity and type II diabetes.ConclusionThe proposed method provides an effective and scalable way for integrating multi omics data to infer disease genes. Further applications of BRIDGE will be benefit to providing novel disease genes and underlying mechanisms of human diseases.

A pathway-based data integration framework for prediction of disease progression

Motivation: Within medical research there is an increasing trend toward deriving multiple types of data from the same individual. The most effective prognostic prediction methods should use all available data, as this maximizes the amount of information used. In this article, we consider a variety of learning strategies to boost prediction performance based on the use of all available data.Implementation: We consider data integration via the use of multiple kernel learning supervised learning methods. We propose a scheme in which feature selection by statistical score is performed separately per data type and by pathway membership. We further consider the introduction of a confidence measure for the class assignment, both to remove some ambiguously labeled datapoints from the training data and to implement a cautious classifier that only makes predictions when the associated confidence is high.Results: We use the METABRIC dataset for breast cancer, with prediction of survival at 2000 days from diagnosis. Predictive accuracy is improved by using kernels that exclusively use those genes, as features, which are known members of particular pathways. We show that yet further improvements can be made by using a range of additional kernels based on clinical covariates such as Estrogen Receptor (ER) status. Using this range of measures to improve prediction performance, we show that the test accuracy on new instances is nearly 80%, though predictions are only made on 69.2% of the patient cohort.Availability: https://github.com/jseoane/FSMKLContact: J.Seoane@bristol.ac.ukSupplementary information: Supplementary data are available at Bioinformatics online.

Microarray characterization of gene expression changes in blood during acute ethanol exposure

BackgroundAs part of the civil aviation safety program to define the adverse effects of ethanol on flying performance, we performed a DNA microarray analysis of human whole blood samples from a five-time point study of subjects administered ethanol orally, followed by breathalyzer analysis, to monitor blood alcohol concentration (BAC) to discover significant gene expression changes in response to the ethanol exposure.MethodsSubjects were administered either orange juice or orange juice with ethanol. Blood samples were taken based on BAC and total RNA was isolated from PaxGene™ blood tubes. The amplified cDNA was used in microarray and quantitative real-time polymerase chain reaction (RT-qPCR) analyses to evaluate differential gene expression. Microarray data was analyzed in a pipeline fashion to summarize and normalize and the results evaluated for relative expression across time points with multiple methods. Candidate genes showing distinctive expression patterns in response to ethanol were clustered by pattern and further analyzed for related function, pathway membership and common transcription factor binding within and across clusters. RT-qPCR was used with representative genes to confirm relative transcript levels across time to those detected in microarrays.ResultsMicroarray analysis of samples representing 0%, 0.04%, 0.08%, return to 0.04%, and 0.02% wt/vol BAC showed that changes in gene expression could be detected across the time course. The expression changes were verified by qRT-PCR.The candidate genes of interest (GOI) identified from the microarray analysis and clustered by expression pattern across the five BAC points showed seven coordinately expressed groups. Analysis showed function-based networks, shared transcription factor binding sites and signaling pathways for members of the clusters. These include hematological functions, innate immunity and inflammation functions, metabolic functions expected of ethanol metabolism, and pancreatic and hepatic function. Five of the seven clusters showed links to the p38 MAPK pathway.ConclusionsThe results of this study provide a first look at changing gene expression patterns in human blood during an acute rise in blood ethanol concentration and its depletion because of metabolism and excretion, and demonstrate that it is possible to detect changes in gene expression using total RNA isolated from whole blood. The analysis approach for this study serves as a workflow to investigate the biology linked to expression changes across a time course and from these changes, to identify target genes that could serve as biomarkers linked to pilot performance.

Pathways of Adult Children Providing Care to Older Parents

Guided by life course and stress process theory, this study investigated pathways of adult child caregivers' family (caregiving, marital, parenting) and nonfamily (employment) roles. Eight waves of data from the Health and Retirement Study were analyzed for 1,300 adult child caregivers. Latent class analysis provided strong evidence for a 4‐class model of caregivers' role pathways. The four pathways were (a) Not‐Married, Early‐Transition to Not‐Working Caregivers (34%), (b) Married, Not‐Working Caregivers (26%), (c) Married, Late‐Transition to Not‐Working Caregivers (23%), and (d) Married, Not‐Working Caregivers with Coresiding Child (17%). Caregivers' background characteristics and contexts predicted pathway membership. Adult child caregivers have structurally diverse life pathways that have implications for theory, research, and practice.

Proteomic and Bioinformatic Profile of Primary Human Oral Epithelial Cells

Wounding of the oral mucosa occurs frequently in a highly septic environment. Remarkably, these wounds heal quickly and the oral cavity, for the most part, remains healthy. Deciphering the normal human oral epithelial cell (NHOEC) proteome is critical for understanding the mechanism(s) of protection elicited when the mucosal barrier is intact, as well as when it is breached. Combining 2D gel electrophoresis with shotgun proteomics resulted in identification of 1662 NHOEC proteins. Proteome annotations were performed based on protein classes, molecular functions, disease association and membership in canonical and metabolic signaling pathways. Comparing the NHOEC proteome with a database of innate immunity-relevant interactions (InnateDB) identified 64 common proteins associated with innate immunity. Comparison with published salivary proteomes revealed that 738/1662 NHOEC proteins were common, suggesting that significant numbers of salivary proteins are of epithelial origin. Gene ontology analysis showed similarities in the distributions of NHOEC and saliva proteomes with regard to biological processes, and molecular functions. We also assessed the interindividual variability of the NHOEC proteome and observed it to be comparable with other primary cells. The baseline proteome described in this study should serve as a resource for proteome studies of the oral mucosa, especially in relation to disease processes.

Linking proteins to signaling pathways for experiment design and evaluation.

Biomedical experimental work often focuses on altering the functions of selected proteins. These changes can hit signaling pathways, and can therefore unexpectedly and non-specifically affect cellular processes. We propose PathwayLinker, an online tool that can provide a first estimate of the possible signaling effects of such changes, e.g., drug or microRNA treatments. PathwayLinker minimizes the users' efforts by integrating protein-protein interaction and signaling pathway data from several sources with statistical significance tests and clear visualization. We demonstrate through three case studies that the developed tool can point out unexpected signaling bias in normal laboratory experiments and identify likely novel signaling proteins among the interactors of known drug targets. In our first case study we show that knockdown of the Caenorhabditis elegans gene cdc-25.1 (meant to avoid progeny) may globally affect the signaling system and unexpectedly bias experiments. In the second case study we evaluate the loss-of-function phenotypes of a less known C. elegans gene to predict its function. In the third case study we analyze GJA1, an anti-cancer drug target protein in human, and predict for this protein novel signaling pathway memberships, which may be sources of side effects. Compared to similar services, a major advantage of PathwayLinker is that it drastically reduces the necessary amount of manual literature searches and can be used without a computational background. PathwayLinker is available at http://PathwayLinker.org. Detailed documentation and source code are available at the website.

Pathway-directed weighted testing procedures for the integrative analysis of gene expression and metabolomic data

We explore the utility of p-value weighting for enhancing the power to detect differential metabolites in a two-sample setting. Related gene expression information is used to assign an a priori importance level to each metabolite being tested. We map the gene expression to a metabolite through pathways and then gene expression information is summarized per-pathway using gene set enrichment tests. Through simulation we explore four styles of enrichment tests and four weight functions to convert the gene information into a meaningful p-value weight. We implement the p-value weighting on a prostate cancer metabolomic dataset. Gene expression on matched samples is used to construct the weights. Under certain regulatory conditions, the use of weighted p-values does not inflate the type I error above what we see for the un-weighted tests except in high correlation situations. The power to detect differential metabolites is notably increased in situations with disjoint pathways and shows moderate improvement, relative to the proportion of enriched pathways, when pathway membership overlaps.

Discriminating response groups in metabolic and regulatory pathway networks

Analysis of omics experiments generates lists of entities (genes, metabolites, etc.) selected based on specific behavior, such as changes in response to stress or other signals. Functional interpretation of these lists often uses category enrichment tests using functional annotations like Gene Ontology terms and pathway membership. This approach does not consider the connected structure of biochemical pathways or the causal directionality of events. The Omics Response Group (ORG) method, described in this work, interprets omics lists in the context of metabolic pathway and regulatory networks using a statistical model for flow within the networks. Statistical results for all response groups are visualized in a novel Pathway Flow plot. The statistical tests are based on the Erlang distribution model under the assumption of independent and identically Exponential-distributed random walk flows through pathways. As a proof of concept, we applied our method to an Escherichia coli transcriptomics dataset where we confirmed common knowledge of the E.coli transcriptional response to Lipid A deprivation. The main response is related to osmotic stress, and we were also able to detect novel responses that are supported by the literature. We also applied our method to an Arabidopsis thaliana expression dataset from an abscisic acid study. In both cases, conventional pathway enrichment tests detected nothing, while our approach discovered biological processes beyond the original studies. We created a prototype for an interactive ORG web tool at http://ecoserver.vrac.iastate.edu/pathwayflow (source code is available from https://subversion.vrac.iastate.edu/Subversion/jlv/public/jlv/pathwayflow). The prototype is described along with additional figures and tables in Supplementary Material. julied@iastate.edu Supplementary data are available at Bioinformatics online.

Identifying and predicting problem behavior trajectories among pre-school children investigated for child abuse and neglect

Objective Young children involved with child welfare services are at high risk for behavior problems. This study aims to identify externalizing behavior paths that preschoolers in this high-risk population follow over a 6 year period, and the predictors of membership in normative and problematic pathways. Methods Using data from the National Study of Child and Adolescent Well-being (NSCAW), the sample included 246 4-year-olds who remained home after investigation. Latent class growth analysis (LCGA) was used to estimate the number, size, and shape of subgroups of preschoolers following distinct behavioral pathways. Early predictors of membership in the resulting groups were then examined. Results Four groups of preschool children following distinct behavior trajectories over 6 years were identified. Weighted results show that more than half (61%) of the children followed a low/normal problem behavior trajectory while just over one tenth (12%) were on a persistent high trajectory, remaining in the clinical range throughout the study. Improving (23%) and worsening (4%) problem behavior groups were also identified. Internalizing problems, attention problems, child ethnicity, and maltreatment type reported at age 4 predicted membership in the trajectory groups. Conclusions Internalizing behavior problems and maltreatment type may distinguish preschool children who are more likely to experience worsening or persistent problematic externalizing behaviors from those likely to follow a normal behavior trajectory. Practice implications Identifying early indicators of externalizing behavior problems and addressing them with evidence-based interventions to reduce negative behaviors may avert long-term negative outcomes.