Large Gene Regulatory Networks Research Articles

SLIVER: Unveiling large scale gene regulatory networks of single-cell transcriptomic data through causal structure learning and modules aggregation

Prevalent Gene Regulatory Network (GRN) construction methods rely on generalized correlation analysis. However, in biological systems, regulation is essentially a causal relationship that cannot be adequately captured solely through correlation. Therefore, it is more reasonable to infer GRNs from a causal perspective. Existing causal discovery algorithms typically rely on Directed Acyclic Graphs (DAGs) to model causal relationships, but it often requires traversing the entire network, which result in computational demands skyrocketing as the number of nodes grows and make causal discovery algorithms only suitable for small networks with one or two hundred nodes or fewer. In this study, we propose the SLIVER (cauSaL dIscovery Via dimEnsionality Reduction) algorithm which integrates causal structural equation model and graph decomposition. SLIVER introduces a set of factor nodes, serving as abstractions of different functional modules to integrate the regulatory relationships between genes based on their respective functions or pathways, thus reducing the GRN to the product of two low-dimensional matrices. Subsequently, we employ the structural causal model (SCM) to learn the GRN within the gene node space, enforce the DAG constraint in the low-dimensional space, and guide each factor to aggregate various functions through cosine similarity. We evaluate the performance of the SLIVER algorithm on 12 real single cell transcriptomic datasets, and demonstrate it outperforms other 12 widely used methods both in GRN inference performance and computational resource usage. The analysis of the gene information integrated by factor nodes also demonstrate the biological explanation of factor nodes in GRNs. We apply it to scRNA-seq of Type 2 diabetes mellitus to capture the transcriptional regulatory structural changes of β cells under high insulin demand.

Network‐based systematic dissection of exercise‐induced inhibition of myosteatosis in older individuals

AbstractAccumulated fat in skeletal muscle (i.e. myosteatosis), common in sedentary older individuals, compromises skeletal muscle health and function. A mechanistic understanding of how physical activity levels dictate fat accumulation represents a critical step towards establishment of therapies that promote healthy ageing. Using a network medicine paradigm that characterized the transcriptomic response of aged muscle to exercise versus immobilization protocols, this study explored the shared molecular cascade that regulates the fate of fibro‐adipogenic progenitors (FAPs), the cell population primarily responsible for fat accumulation. Specifically, gene set enrichment analyses with network propagation revealed Pgc‐1α as a functional hub of a large gene regulatory network underlying the regulation of FAPs by physical activity in aged muscle, but not in young counterparts. Integrated in silico and in situ approaches to induce Pgc‐1α overexpression in aged muscle promoted mitochondrial fatty acid oxidation and inhibited FAP adipogenesis. These findings suggest that the Pgc‐1α–mitochondrial fatty acid oxidation axis is a shared mechanism by which physical activity regulates age‐related myosteatosis. The network medicine paradigm introduced provides mechanistic insight into exercise adaptation in elderly skeletal muscle and offers translational opportunities to advance exercise prescription for older populations. imageKey points Fat accumulation is a quintessential feature of aged skeletal muscle. While increasing physical activity levels has been proposed as an effective strategy to reduce the fat in skeletal muscle (i.e. myosteatosis), the molecular cascade underlying these benefits has been poorly defined. This study implemented a series of network medicine approaches and uncovered Pgc‐1α as a mechanistic driver of the regulation of fibro‐adipogenic progenitors (FAPs) by physical activity. Integrated in silico and in situ approaches to induce Pgc‐1α overexpression promoted mitochondrial fatty acid oxidation and inhibited FAP adipogenesis. Together, the findings of the current study suggest a novel hypothesis that physical activity reduces myosteatosis via upregulation of Pgc‐1α‐mediated mitochondrial fatty acid oxidation and subsequent inhibition of FAP adipogenesis.

High-resolution Nanopore methylome-maps reveal random hyper-methylation at CpG-poor regions as driver of chemoresistance in leukemias

Aberrant DNA methylation at CpG dinucleotides is a cancer hallmark that is associated with the emergence of resistance to anti cancer treatment, though molecular mechanisms and biological significance remain elusive. Genome scale methylation maps by currently used methods are based on chemical modification of DNA and are best suited for analyses of methylation at CpG rich regions (CpG islands). We report the first high coverage whole-genome map in cancer using the long read nanopore technology, which allows simultaneous DNA-sequence and -methylation analyses on native DNA. We analyzed clonal epigenomic/genomic evolution in Acute Myeloid Leukemias (AMLs) at diagnosis and relapse, after chemotherapy. Long read sequencing coupled to a novel computational method allowed definition of differential methylation at unprecedented resolution, and showed that the relapse methylome is characterized by hypermethylation at both CpG islands and sparse CpGs regions. Most differentially methylated genes, however, were not differentially expressed nor enriched for chemoresistance genes. A small fraction of under-expressed and hyper-methylated genes at sparse CpGs, in the gene body, was significantly enriched in transcription factors (TFs). Remarkably, these few TFs supported large gene-regulatory networks including 50% of all differentially expressed genes in the relapsed AMLs and highly-enriched in chemoresistance genes. Notably, hypermethylated regions at sparse CpGs were poorly conserved in the relapsed AMLs, under-represented at their genomic positions and showed higher methylation entropy, as compared to CpG islands. Analyses of available datasets confirmed TF binding to their target genes and conservation of the same gene-regulatory networks in large patient cohorts. Relapsed AMLs carried few patient specific structural variants and DNA mutations, apparently not involved in drug resistance. Thus, drug resistance in AMLs can be mainly ascribed to the selection of random epigenetic alterations at sparse CpGs of a few transcription factors, which then induce reprogramming of the relapsing phenotype, independently of clonal genomic evolution.

Regulatory Networks of LncRNA MALAT-1 in Cancer.

Long noncoding (lnc)RNAs are a group of RNAs with a length greater than 200 nt that do not encode a protein but play an essential role in regulating the expression of target genes in normal biological contexts as well as pathologic processes including tumorigenesis. The lncRNA metastasis-associated lung adenocarcinoma transcript (MALAT)-1 has been widely studied in cancer. In this review, we describe the known functions of MALAT-1; its mechanisms of action; and associated signaling pathways and their clinical significance in different cancers. In most malignancies, including lung, colorectal, thyroid, and other cancers, MALAT-1 functions as an oncogene and is upregulated in tumors and tumor cell lines. MALAT-1 has a distinct mechanism of action in each cancer type and is thus at the center of large gene regulatory networks. Dysregulation of MALAT-1 affects cellular processes such as alternative splicing, epithelial–mesenchymal transition, apoptosis, and autophagy, which ultimately results in the abnormal cell proliferation, invasion, and migration that characterize cancers. In other malignancies, such as glioma and endometrial carcinoma, MALAT-1 functions as a tumor suppressor and thus forms additional regulatory networks. The current evidence indicates that MALAT-1 and its associated signaling pathways can serve as diagnostic or prognostic biomarker or therapeutic target in the treatment of many cancers.

Identification of a new set of drought-related miRNA-SSR markers and association analysis under drought stress in rice (Oryza sativa L.)

The miRNAs are a part of large gene regulatory networks, and the identification of molecular markers linked to miRNA genes can be useful for breeding programs. In this study, 10 microsatellite markers were mined from drought-responsive miRNA genes in rice. Based on the analysis of genetic diversity, a total of seven polymorphic miRNA-SSR markers from among 83 diverse rice genotypes showed high levels of genetic diversity. The structure analysis based on 34 SSR markers and seven miRNA-SSR markers divided the genotypes into two distinct groups. The association analysis between seven miRNA-SSR markers and 21 phenotypic traits using the population structure data as covariates showed five pairs of associations at the significant levels of P ≤ .01 when tested using the two statistical models. In this study, we demonstrated a relation between miRNA-SSR markers and some quantitative traits, and also defined useful consistent linked markers in reproductive stress and non-stress conditions.

Connecting the molecular function of microRNAs to cell differentiation dynamics.

MicroRNAs form a class of short, non-coding RNA molecules which are essential for proper development in tissue-based plants and animals. To help explain their role in gene regulation, a number of mathematical and computational studies have demonstrated the potential canalizing effects of microRNAs. However, such studies have typically focused on the effects of microRNAs on only one or a few target genes. Consequently, it remains unclear how these small-scale effects add up to the experimentally observed developmental outcomes resulting from microRNA perturbation at the whole-genome level. To answer this question, we built a general computational model of cell differentiation to study the effect of microRNAs in genome-scale gene regulatory networks. Our experiments show that in large gene regulatory networks, microRNAs can control differentiation time without significantly changing steady-state gene expression profiles. This temporal regulatory role cannot be naturally replicated using protein-based transcription factors alone. While several microRNAs have been shown to regulate differentiation time in vivo, our findings provide a new explanation of how the cumulative molecular actions of individual microRNAs influence genome-scale cellular dynamics. Taken together, these results may help explain why tissue-based organisms exclusively depend on miRNA-mediated regulation, while their more primitive counterparts do not.

A stochastic switch with different phases.

We describe an analog stochastic switch that exhibits three distinct phases as its parameters change. The phases are classified by the mean and variance of the switch's output. A phase change appears if the mean or the variance tends to a finite value or to infinity. The switch can be embedded in a large gene regulatory network for which the moment equations naturally close at the second order. This switch was used to model the response of a heat-shock system.

Scalable optimal Bayesian classification of single-cell trajectories under regulatory model uncertainty

BackgroundSingle-cell gene expression measurements offer opportunities in deriving mechanistic understanding of complex diseases, including cancer. However, due to the complex regulatory machinery of the cell, gene regulatory network (GRN) model inference based on such data still manifests significant uncertainty.ResultsThe goal of this paper is to develop optimal classification of single-cell trajectories accounting for potential model uncertainty. Partially-observed Boolean dynamical systems (POBDS) are used for modeling gene regulatory networks observed through noisy gene-expression data. We derive the exact optimal Bayesian classifier (OBC) for binary classification of single-cell trajectories. The application of the OBC becomes impractical for large GRNs, due to computational and memory requirements. To address this, we introduce a particle-based single-cell classification method that is highly scalable for large GRNs with much lower complexity than the optimal solution.ConclusionThe performance of the proposed particle-based method is demonstrated through numerical experiments using a POBDS model of the well-known T-cell large granular lymphocyte (T-LGL) leukemia network with noisy time-series gene-expression data.

Adaptive Particle Filtering for Fault Detection in Partially-Observed Boolean Dynamical Systems

We propose a novel methodology for fault detection and diagnosis in partially-observed Boolean dynamical systems (POBDS). These are stochastic, highly nonlinear, and derivativeless systems, rendering difficult the application of classical fault detection and diagnosis methods. The methodology comprises two main approaches. The first addresses the case when the normal mode of operation is known but not the fault modes. It applies an innovations filter (IF) to detect deviations from the nominal normal mode of operation. The second approach is applicable when the set of possible fault models is finite and known, in which case we employ a multiple model adaptive estimation (MMAE) approach based on a likelihood-ratio (LR) statistic. Unknown system parameters are estimated by an adaptive expectation-maximization (EM) algorithm. Particle filtering techniques are used to reduce the computational complexity in the case of systems with large state-spaces. The efficacy of the proposed methodology is demonstrated by numerical experiments with a large gene regulatory network (GRN) with stuck-at faults observed through a single noisy time series of RNA-seq gene expression measurements.

Deterministic and Stochastic Models of Arabidopsis thaliana Flowering

Experimental studies of the flowering of Arabidopsis thaliana have shown that a large complex gene regulatory network (GRN) is responsible for its regulation. This process has been mathematically modelled with deterministic differential equations by considering the interactions between gene activators and inhibitors (Valentim et al. in PLoS ONE 10(2):e0116973, 2015; van Mourik et al. in BMC Syst Biol 4(1):1, 2010). However, due to complexity of the model, the properties of the network and the roles of the individual genes cannot be deducted from the numerical solution the published work offers. Here, we propose simplifications of the model, based on decoupling of the original GRN to motifs, described with three and two differential equations. A stable solution of the original model is sought by linearisation of the original model which contributes to further investigation of the role of the individual genes to the flowering. Furthermore, we study the role of noise by introducing and investigating two types of stochastic elements into the model. The deterministic and stochastic nonlinear dynamic models of Arabidopsis flowering time are considered by following the deterministic delayed model introduced in Valentim et al. (2015). Steady-state regimes and stability of the deterministic original model are investigated analytically and numerically. By decoupling some concentrations, the system was reduced to emphasise the role played by the transcription factor Suppressor of Overexpression of Constants1 (textit{SOC}1) and the important floral meristem identity genes, Leafy (textit{LFY}) and Apetala1 (textit{AP}1). Two-dimensional motifs, based on the dynamics of textit{LFY} and textit{AP}1, are obtained from the reduced network and parameter ranges ensuring flowering are determined. Their stability analysis shows that textit{LFY} and textit{AP}1 are regulating each other for flowering, matching experimental findings. New sufficient conditions of mean square stability in the stochastic model are obtained using a stochastic Lyapunov approach. Our numerical simulations demonstrate that the reduced models of Arabidopsis flowering time, describing specific motifs of the GRN, can capture the essential behaviour of the full system and also introduce the conditions of flowering initiation. Additionally, they show that stochastic effects can change the behaviour of the stability region through a stability switch. This study thus contributes to a better understanding of the role of textit{LFY} and textit{AP}1 in Arabidopsis flowering.

Decomposition Approach for Learning Large Gene Regulatory Network

Gene Regulatory Network (GRN) represents the complex interaction between Transcription Factors (TFs) and other genes with time delays. They are important in the working of the cell. Learning GRN is an important first step towards understanding the working of the cell and consequently curing diseases related to malfunctioning of the cell. One significant problem in learning GRN is that the available time series expression data is still limited compared to the network size. To alleviate this problem, besides using multiple expression replicates, we propose to decompose large network into small subnetwork without prior knowledge. Our algorithm first infers an initial GRN using CLINDE, then decomposes it into possibly overlapping subnetworks, then infers each subnetwork by either CLINDE or DD-lasso and finally merges the subnetworks. We have tested this algorithm on synthetic data of many networks with 500 and 1000 genes. We have also tested on real data on 41 human TF regulatory networks. Results show that our proposed algorithm does improve the GRN learning performance of using either CLINDE or DD-lasso alone on the large network.

Wound healing, calcium signaling, and other novel pathways are associated with the formation of butterfly eyespots

BackgroundOne hypothesis surrounding the origin of novel traits is that they originate from the co-option of pre-existing genes or larger gene regulatory networks into novel developmental contexts. Insights into a trait’s evolutionary origins can, thus, be gained via identification of the genes underlying trait development, and exploring whether those genes also function in other developmental contexts. Here we investigate the set of genes associated with the development of eyespot color patterns, a trait that originated once within the Nymphalid family of butterflies. Although several genes associated with eyespot development have been identified, the eyespot gene regulatory network remains largely unknown.ResultsIn this study, next-generation sequencing and transcriptome analyses were used to identify a large set of genes associated with eyespot development of Bicyclus anynana butterflies, at 3-6 h after pupation, prior to the differentiation of the color rings. Eyespot-associated genes were identified by comparing the transcriptomes of homologous micro-dissected wing tissues that either develop or do not develop eyespots in wild-type and a mutant line of butterflies, Spotty, with extra eyespots. Overall, 186 genes were significantly up and down-regulated in wing tissues that develop eyespots compared to wing tissues that do not. Many of the differentially expressed genes have yet to be annotated. New signaling pathways, including the Toll, Fibroblast Growth Factor (FGF), extracellular signal–regulated kinase (ERK) and/or Jun N-terminal kinase (JNK) signaling pathways are associated for the first time with eyespot development. In addition, several genes involved in wound healing and calcium signaling were also found to be associated with eyespots.ConclusionsOverall, this study provides the identity of many new genes and signaling pathways associated with eyespots, and suggests that the ancient wound healing gene regulatory network may have been co-opted to cells at the center of the pattern to aid in eyespot origins. New transcription factors that may be providing different identities to distinct wing sectors, and genes with sexually dimorphic expression in the eyespots were also identified.

Identifying genes and regulatory pathways associated with the scleractinian coral calcification process.

Reef building corals precipitate calcium carbonate as an exo-skeleton and provide substratum for prosperous marine life. Biomineralization of the coral’s skeleton is a developmental process that occurs concurrently with other proliferation processes that control the animal extension and growth. The development of the animal body is regulated by large gene regulatory networks, which control the expression of gene sets that progressively generate developmental patterns in the animal body. In this study we have explored the gene expression profile and signaling pathways followed by the calcification process of a basal metazoan, the Red Sea scleractinian (stony) coral, Stylophora pistillata. When treated by seawater with high calcium concentrations (addition of 100 gm/L, added as CaCl2.2H2O), the coral increases its calcification rates and associated genes were up-regulated as a result, which were then identified. Gene expression was compared between corals treated with elevated and normal calcium concentrations. Calcification rate measurements and gene expression analysis by microarray RNA transcriptional profiling at two time-points (midday and night-time) revealed several genes common within mammalian gene regulatory networks. This study indicates that core genes of the Wnt and TGF-β/BMP signaling pathways may also play roles in development, growth, and biomineralization in early-diverging organisms such as corals.

Reverse engineering highlights potential principles of large gene regulatory network design and learning

Inferring transcriptional gene regulatory networks from transcriptomic datasets is a key challenge of systems biology, with potential impacts ranging from medicine to agronomy. There are several techniques used presently to experimentally assay transcription factors to target relationships, defining important information about real gene regulatory networks connections. These techniques include classical ChIP-seq, yeast one-hybrid, or more recently, DAP-seq or target technologies. These techniques are usually used to validate algorithm predictions. Here, we developed a reverse engineering approach based on mathematical and computer simulation to evaluate the impact that this prior knowledge on gene regulatory networks may have on training machine learning algorithms. First, we developed a gene regulatory networks-simulating engine called FRANK (Fast Randomizing Algorithm for Network Knowledge) that is able to simulate large gene regulatory networks (containing 104 genes) with characteristics of gene regulatory networks observed in vivo. FRANK also generates stable or oscillatory gene expression directly produced by the simulated gene regulatory networks. The development of FRANK leads to important general conclusions concerning the design of large and stable gene regulatory networks harboring scale free properties (built ex nihilo). In combination with supervised (accepting prior knowledge) support vector machine algorithm we (i) address biologically oriented questions concerning our capacity to accurately reconstruct gene regulatory networks and in particular we demonstrate that prior-knowledge structure is crucial for accurate learning, and (ii) draw conclusions to inform experimental design to performed learning able to solve gene regulatory networks in the future. By demonstrating that our predictions concerning the influence of the prior-knowledge structure on support vector machine learning capacity holds true on real data (Escherichia coli K14 network reconstruction using network and transcriptomic data), we show that the formalism used to build FRANK can to some extent be a reasonable model for gene regulatory networks in real cells.

A Generalized Gene-Regulatory Network Model of Stem Cell Differentiation for Predicting Lineage Specifiers.

SummaryIdentification of cell-fate determinants for directing stem cell differentiation remains a challenge. Moreover, little is known about how cell-fate determinants are regulated in functionally important subnetworks in large gene-regulatory networks (i.e., GRN motifs). Here we propose a model of stem cell differentiation in which cell-fate determinants work synergistically to determine different cellular identities, and reside in a class of GRN motifs known as feedback loops. Based on this model, we develop a computational method that can systematically predict cell-fate determinants and their GRN motifs. The method was able to recapitulate experimentally validated cell-fate determinants, and validation of two predicted cell-fate determinants confirmed that overexpression of ESR1 and RUNX2 in mouse neural stem cells induces neuronal and astrocyte differentiation, respectively. Thus, the presented GRN-based model of stem cell differentiation and computational method can guide differentiation experiments in stem cell research and regenerative medicine.

"Hit-and-Run" transcription: de novo transcription initiated by a transient bZIP1 "hit" persists after the "run".

BackgroundDynamic transcriptional regulation is critical for an organism’s response to environmental signals and yet remains elusive to capture. Such transcriptional regulation is mediated by master transcription factors (TF) that control large gene regulatory networks. Recently, we described a dynamic mode of TF regulation named “hit-and-run”. This model proposes that master TF can interact transiently with a set of targets, but the transcription of these transient targets continues after the TF dissociation from the target promoter. However, experimental evidence validating active transcription of the transient TF-targets is still lacking.ResultsHere, we show that active transcription continues after transient TF-target interactions by tracking de novo synthesis of RNAs made in response to TF nuclear import. To do this, we introduced an affinity-labeled 4-thiouracil (4tU) nucleobase to specifically isolate newly synthesized transcripts following conditional TF nuclear import. Thus, we extended the TARGET system (Transient Assay Reporting Genome-wide Effects of Transcription factors) to include 4tU-labeling and named this new technology TARGET-tU. Our proof-of-principle example is the master TF Basic Leucine Zipper 1 (bZIP1), a central integrator of metabolic signaling in plants. Using TARGET-tU, we captured newly synthesized mRNAs made in response to bZIP1 nuclear import at a time when bZIP1 is no longer detectably bound to its target. Thus, the analysis of de novo transcripomics demonstrates that bZIP1 may act as a catalyst TF to initiate a transcriptional complex (“hit”), after which active transcription by RNA polymerase continues without the TF being bound to the gene promoter (“run”).ConclusionOur findings provide experimental proof for active transcription of transient TF-targets supporting a “hit-and-run” mode of action. This dynamic regulatory model allows a master TF to catalytically propagate rapid and broad transcriptional responses to changes in environment. Thus, the functional read-out of de novo transcripts produced by transient TF-target interactions allowed us to capture new models for genome-wide transcriptional control.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-2410-2) contains supplementary material, which is available to authorized users.

Large scale gene regulatory network inference with a multi-level strategy.

Transcriptional regulation is a basis of many crucial molecular processes and an accurate inference of the gene regulatory network is a helpful and essential task to understand cell functions and gain insights into biological processes of interest in systems biology. Inspired by the Dialogue for Reverse Engineering Assessments and Methods (DREAM) projects, many excellent gene regulatory network inference algorithms have been proposed. However, it is still a challenging problem to infer a gene regulatory network from gene expression data on a large scale. In this paper, we propose a gene regulatory network inference method based on a multi-level strategy (GENIMS), which can give results that are more accurate and robust than the state-of-the-art methods. The proposed method mainly consists of three levels, which are an original feature selection step based on guided regularized random forest, normalization of individual feature selection and the final refinement step according to the topological property of the gene regulatory network. To prove the accuracy and robustness of our method, we compare our method with the state-of-the-art methods on the DREAM4 and DREAM5 benchmark networks and the results indicate that the proposed method can significantly improve the performance of gene regulatory network inference. Additionally, we also discuss the influence of the selection of different parameters in our method.

Are Quasi-Steady-State Approximated Models Suitable for Quantifying Intrinsic Noise Accurately?

Large gene regulatory networks (GRN) are often modeled with quasi-steady-state approximation (QSSA) to reduce the huge computational time required for intrinsic noise quantification using Gillespie stochastic simulation algorithm (SSA). However, the question still remains whether the stochastic QSSA model measures the intrinsic noise as accurately as the SSA performed for a detailed mechanistic model or not? To address this issue, we have constructed mechanistic and QSSA models for few frequently observed GRNs exhibiting switching behavior and performed stochastic simulations with them. Our results strongly suggest that the performance of a stochastic QSSA model in comparison to SSA performed for a mechanistic model critically relies on the absolute values of the mRNA and protein half-lives involved in the corresponding GRN. The extent of accuracy level achieved by the stochastic QSSA model calculations will depend on the level of bursting frequency generated due to the absolute value of the half-life of either mRNA or protein or for both the species. For the GRNs considered, the stochastic QSSA quantifies the intrinsic noise at the protein level with greater accuracy and for larger combinations of half-life values of mRNA and protein, whereas in case of mRNA the satisfactory accuracy level can only be reached for limited combinations of absolute values of half-lives. Further, we have clearly demonstrated that the abundance levels of mRNA and protein hardly matter for such comparison between QSSA and mechanistic models. Based on our findings, we conclude that QSSA model can be a good choice for evaluating intrinsic noise for other GRNs as well, provided we make a rational choice based on experimental half-life values available in literature.

Large Causal Gene Regulatory Network Inference by Decomposition into Subnetworks

Large Causal Gene Regulatory Network Inference by Decomposition into Subnetworks

Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes

Lymphoma is one of the most common malignancies in dogs. Canine lymphoma is similar to human non-Hodgkin's lymphoma (NHL) with shared clinical presentation and histopathological features. This study reports the construction of a comprehensive gene regulatory network (GRN) for canine diffuse large B-cell lymphoma (DLBCL), the most common type of canine lymphoma, and performs analysis for detection of major functional modules and hub genes (the most important genes in a GRN). The canine DLBCL GRN was reconstructed from gene expression data (NCBI GEO dataset: GSE30881) using the STRING and MiMI interaction databases. Reconstructed GRNs were then assessed, using various bioinformatics programmes, in order to analyze network topology and identify major pathways and hub genes. The resultant network from both interaction databases had a logically scale-free pattern. Gene ontology (GO) analysis revealed cell activation, cell cycle phase, immune effector process, immune system development, immune system process, integrin-mediated signalling pathway, intracellular protein kinase cascade, intracellular signal transduction, leucocyte activation and differentiation, lymphocyte activation and differentiation as major GO terms in the biological processes of the networks. Moreover, bioinformatics analysis showed E2F1, E2F4, PTEN, CDKN1A, PCNA, DKC1, MNAT1, NDUFB4, ATP5J, PRKDC, BRCA1, MYCN, RFC4 and POLA1 as the most important hub genes. The phosphatidyl inositol signalling system, P53 signalling pathway, Rac CycD pathway, G1/S checkpoint, chemokine signalling pathway and telomere maintenance were the main signalling pathways in which the protein products of the hub genes are involved.