Search

Bigenic expression relationships are conventionally defined based on metrics such as Pearson or Spearman correlation that cannot typically detect latent, non-linear dependencies or require the relationship to be monotonic. Further, the combination of intrinsic and extrinsic noise as well as embedded relationships between sample sub-populations reduces the probability of extracting biologically relevant edges during the construction of gene co-expression networks (GCNs). In this report, we address these problems via our NetExtractor algorithm. NetExtractor examines all pairwise gene expression profiles first with Gaussian mixture models (GMMs) to identify sample sub-populations followed by mutual information (MI) analysis that is capable of detecting non-linear differential bigenic expression relationships. We applied NetExtractor to brain tissue RNA profiles from the Genotype-Tissue Expression (GTEx) project to obtain a brain tissue specific gene expression relationship network centered on cerebellar and cerebellar hemisphere enriched edges. We leveraged the PsychENCODE pre-frontal cortex (PFC) gene regulatory network (GRN) to construct a cerebellar cortex (cerebellar) GRN associated with transcriptionally active regions in cerebellar tissue. Thus, we demonstrate the utility of our NetExtractor approach to detect biologically relevant and novel non-linear binary gene relationships.

The objective of this study was to assess how exposure to ergot alkaloids during 2 stages of gestation alters fetal growth, muscle fiber formation, and miRNA expression. Pregnant ewes (n = 36; BW = 83.26 ± 8.14 kg; 4/group; 9 groups) were used in a 2 × 2 factorial arrangement with 2 tall fescue seed treatments [endophyte-infected (E+) vs. endophyte-free (E−)] fed during 2 stages of gestation (MID, days 35 to 85 vs. LATE, days 86 to 133), which created 4 possible treatments (E−/E−, E+/E−, E−/E+, or E+/E+). Ewes were individually fed a total mixed ration containing E+ or E− fescue seed according to treatment assignment. Terminal surgeries were conducted on day 133 of gestation for the collection of fetal measurements and muscle samples. Data were analyzed as a 2 × 2 factorial with fescue treatment, stage of gestation, and 2-way interaction as fixed effects. Fetuses exposed to E+ seed during LATE gestation had reduced (P = 0.0020) fetal BW by 10% compared with E− fetuses; however, fetal body weight did not differ (P = 0.41) with E+ exposure during MID gestation. Fetuses from ewes fed E+ seed during MID and LATE gestation tended to have smaller (P = 0.058) kidney weights compared with E− fetuses. Liver weight was larger (P = 0.0069) in fetuses fed E− during LATE gestation compared with E+. Fetal brain weight did not differ by fescue treatment fed during MID (P = 0.36) or LATE (P = 0.40) gestation. The percentage of brain to empty body weight (EBW) was greater (P = 0.0048) in fetuses from ewes fed E+ fescue seed during LATE gestation, which is indicative of intrauterine growth restriction (IUGR). Primary muscle fiber number was lower (P = 0.0005) in semitendinosus (STN) of fetuses exposed to E+ during MID and/or LATE gestation compared with E−/E−. miRNA sequencing showed differential expression (P < 0.010) of 6 novel miRNAs including bta-miR-652_R+1, mdo-miR-22-3p, bta-miR-1277_R-1, ppy-miR-133a_L+1_1ss5TG, hsa-miR-129-1-3p, and ssc-miR-615 in fetal STN muscle. These miRNA are associated with glucose transport, insulin signaling, intracellular ATP, hypertension, or adipogenesis. This work supports the hypothesis that E+ tall fescue seed fed during late gestation reduces fetal weight and causes asymmetrical growth, which is indicative of IUGR. Changes in primary fiber number and miRNA of STN indicate that exposure to E+ fescue fed during MID and LATE gestation alters fetal muscle development that may affect postnatal muscle growth and meat quality.

The human brain is a complex organ that consists of several regions each with a unique gene expression pattern. Our intent in this study was to construct a gene co-expression network (GCN) for the normal brain using RNA expression profiles from the Genotype-Tissue Expression (GTEx) project. The brain GCN contains gene correlation relationships that are broadly present in the brain or specific to thirteen brain regions, which we later combined into six overarching brain mini-GCNs based on the brain’s structure. Using the expression profiles of brain region-specific GCN edges, we determined how well the brain region samples could be discriminated from each other, visually with t-SNE plots or quantitatively with the Gene Oracle deep learning classifier. Next, we tested these gene sets on their relevance to human tumors of brain and non-brain origin. Interestingly, we found that genes in the six brain mini-GCNs showed markedly higher mutation rates in tumors relative to matched sets of random genes. Further, we found that cortex genes subdivided Head and Neck Squamous Cell Carcinoma (HNSC) tumors and Pheochromocytoma and Paraganglioma (PCPG) tumors into distinct groups. The brain GCN and mini-GCNs are useful resources for the classification of brain regions and identification of biomarker genes for brain related phenotypes.

Renal cell carcinoma (RCC) subtypes are characterized by distinct molecular profiles. Using RNA expression profiles from 1,009 RCC samples, we constructed a condition-annotated gene coexpression network (GCN). The RCC GCN contains binary gene coexpression relationships (edges) specific to conditions including RCC subtype and tumor stage. As an application of this resource, we discovered RCC GCN edges and modules that were associated with genetic lesions in known RCC driver genes, including VHL, a common initiating clear cell RCC (ccRCC) genetic lesion, and PBRM1 and BAP1 which are early genetic lesions in the Braided Cancer River Model (BCRM). Since ccRCC tumors with PBRM1 mutations respond to targeted therapy differently than tumors with BAP1 mutations, we focused on ccRCC-specific edges associated with tumors that exhibit alternate mutation profiles: VHL-PBRM1 or VHL-BAP1. We found specific blends molecular functions associated with these two mutation paths. Despite these mutation-associated edges having unique genes, they were enriched for the same immunological functions suggesting a convergent functional role for alternate gene sets consistent with the BCRM. The condition annotated RCC GCN described herein is a novel data mining resource for the assignment of polygenic biomarkers and their relationships to RCC tumors with specific molecular and mutational profiles.

BackgroundLinking genotype to phenotype is a major aim of genetics research, yet many complex conditions continue to hide their underlying biochemical mechanisms. Recent research provides evidence that relevant gene-phenotype associations are discoverable in the study of intellectual disability (ID). Here we expand on that work, identifying distinctive gene interaction modules with unique enrichment patterns reflective of associated clinical features in ID.MethodsTwo hundred twelve forms of monogenic ID were curated according to comorbidities with autism and epilepsy. These groups were further subdivided according to secondary clinical symptoms of complex versus simple facial dysmorphia and neurodegenerative-like features due to their clinical prominence, modest symptom overlap, and probable etiological divergence. An aggregate gene interaction ID network for these phenotype subgroups was discovered using via a public database of known gene interactions: protein-protein, genetic, and mRNA coexpression. Additional annotation resources (Gene Ontology, Human Phenotype Ontology, TRANSFAC/JASPAR, and KEGG/WikiPathways) were utilized to assess functional and phenotypic enrichment modules within the full ID network.ResultsPhenotypic analysis revealed high rates of complex facial dysmorphia in ID with comorbid autism. In contrast, neurodegenerative-like features were overrepresented in ID with epilepsy. Network analysis subsequently showed that gene groups divided according to clinical features of interest resulted in distinctive interaction clusters, with unique functional enrichments according to module.ConclusionsThese data suggest that specific comorbid and secondary clinical features in ID are predictive of underlying genotype. In summary, ID form unique clusters, which are comprised of individual conditions with remarkable genotypic and phenotypic overlap.

A gene co-expression network (GCN) describes associations between genes and points to genetic coordination of biochemical pathways. However, genetic correlations in a GCN are only detectable if they are present in the sampled conditions. With the increasing quantity of gene expression samples available in public repositories, there is greater potential for discovery of genetic correlations from a variety of biologically interesting conditions. However, even if gene correlations are present, their discovery can be masked by noise. Noise is introduced from natural variation (intrinsic and extrinsic), systematic variation (caused by sample measurement protocols and instruments), and algorithmic and statistical variation created by selection of data processing tools. A variety of published studies, approaches and methods attempt to address each of these contributions of variation to reduce noise. Here we describe an approach using Gaussian Mixture Models (GMMs) to address natural extrinsic (condition-specific) variation during network construction from mixed input conditions. To demonstrate utility, we build and analyze a condition-annotated GCN from a compendium of 2,016 mixed gene expression data sets from five tumor subtypes obtained from The Cancer Genome Atlas. Our results show that GMMs help discover tumor subtype specific gene co-expression patterns (modules) that are significantly enriched for clinical attributes.

Glucocorticoids indirectly alter adrenocortical steroid output through the inhibition of ACTH secretion by the anterior pituitary. However, previous studies suggest that glucocorticoids can directly affect adrenocortical steroid production. Therefore, we have investigated the ability of glucocorticoids to affect transcription of adrenocortical steroid biosynthetic enzymes. One potential target of glucocorticoid action in the adrenal is an enzyme critical for adrenocortical steroid production: 3beta-hydroxysteroid dehydrogenase/Delta5-Delta4 isomerase (3beta-HSD). Treatment of the adrenocortical cell line (H295R) with the glucocorticoid agonist dexamethasone (DEX) increased cortisol production and 3beta-HSD mRNA levels alone or in conjunction with phorbol ester. This increase in 3beta-HSD mRNA was paralleled by increases in Steroidogenic Acute Regulatory Protein (StAR) mRNA levels. The human type II 3beta-HSD promoter lacks a consensus palindromic glucocorticoid response element (GRE) but does contain a Stat5 response element (Stat5RE) suggesting that glucocorticoids could affect type II 3beta-HSD transcription via interaction with Stat5. Transfection experiments show enhancement of human type II 3beta-HSD promoter activity by coexpression of the glucocorticoid receptor (GR) and Stat5A and treatment with 100nM dexamethasone. Furthermore, removal of the Stat5RE either by truncation of the 5' flanking sequence in the promoter or introduction of point mutations to the Stat5RE abolished the ability of DEX to enhance 3beta-HSD promoter activity. These studies demonstrate the ability of glucocorticoids to directly enhance the expression of an adrenal steroidogenic enzyme gene albeit independent of a consensus palindromic glucocorticoid response element.

Identification of genes and pathways involving in diseases and physiological conditions is a major task in systems biology. In this study, we develop a new non-parameter Ising model to integrate protein-protein interaction network and microarray data for identifying differentially expressed (DE) genes. We also propose a simulated annealing algorithm to find the optimal configuration of the Ising model. We test the Ising model to two breast cancer microarray data sets. The results show that more cancer related differentially expressed subnetworks and genes are identified by the Ising model than by the Markov random filed (MRF) model.

Premature ovarian failure (POF) is the loss of ovarian function in women less than 40 yr of age (reviewed in Ref. 1). It is associated with sex steroid deficiency, amenorrhea, infertility, and elevated serum gonadotropins. While there are multiple etiologies of POF including the exposure to iatrogenic treatments (chemotherapy, radiation), viral agents, and rare genetic disorders, in most patients no etiology can be identified (idiopathic POF). Significant evidence suggests that autoimmunity is a cause of some forms of ovarian failure although specific ovarian antigens are not known and the mechanisms of autoimmune disease development are unclear. Autoimmune POF in humans is frequently associated with other manifestations of autoimmune disease. For example, POF can precede the onset of Addison’s disease or adrenal autoimmunity leading to a deficiency of adrenocortical hormones (1). Autoimmune POF is characterized by inflammatory infiltration of developing follicles, production of antiovarian antibodies, atrophy, and sparing of primordial follicles (1‐3). Autoantibodies in these diseases sometimes react with common antigens in steroid-producing cells of the ovary and adrenal cortex. Common antigens identified have been steroidogenic enzymes including P450 side-chain cleavage (P450scc), 17a-hydroxylase, and 3b-hydroxysteroid dehydrogenase (4 ‐ 6). The identification of specific antigens involved in POF is important for multiple reasons. First, the development of appropriate reagents to screen for the presence of antibodies to these antigens could provide an analytical tool for diagnosing the disease, identifying patients at risk for developing the disease, and detecting patients who may respond to immune-modulating therapies. Second, these tools could be used in research to further understand the mechanisms of disease development and the mechanisms of ovarian pathology associated with the disease. Lastly, the identification of these antigens provides new information on novel proteins in the ovary and their potential function. Animal models of autoimmune premature ovarian failure have yielded important insight into both potential mechanisms of autoimmune disease development and ovarian antigens that may affect disease progression. These models for autoimmune ovarian failure can be induced by multiple methods such as immunization with specific ovarian antigens or neonatal thymectomy in specific genetic strains of mice. A detailed review of the findings will not be repeated here but some major developments will be summarized. The most important development in these animal models of autoimmune ovarian failure comes from multiple studies, all suggesting that the basis of the disease is a cell-mediated autoimmune reaction caused by an alteration in T cell regulation (1, 3). This is most evident in the neonatal thymectomy animal model. The removal of the thymus in specific genetic strains of mice (e.g. BALB/c or A/J) between postnatal days 2 and 5 results in autoimmune ovarian failure. There is a progressive onset of the disease that is potentiated by puberty and the most severe inflammation occurs between 4 ‐14 weeks after thymectomy (3). The proposed mechanism of the disease (3) is that autoreactive T cells (CD41) are generated during normal processes such as apoptosis of follicles in the ovary. These autoreactive cells are normally controlled by CD41 T cells with suppressor activity. However, because these cells are generated in the thymus after the first week of life, neonatal thymectomy results in a dramatic loss in T cells with suppressor function. This animal model strongly implicates T cell regulation in the disease process.