Models Of Regulatory Networks Research Articles

BindCompare: A Novel Integrated Protein-Nucleic Acid Binding Analysis Platform.

Advanced genomic technologies have generated thousands of protein-nucleic acid binding datasets that have the potential to identify testable gene regulatory network (GRNs) models governed by combinatorial associations between factors. Transcription factors (TFs) and RNA binding proteins (RBPs) are nucleic-acid binding proteins regulating gene expression and are key drivers of GRN function. However, the combinatorial mechanisms by which the interactions between specific TFs and RBPs regulate gene expression remain largely unknown. To identify possible combinations of TFs and RBPs that may function together, developing a tool that compares and contrasts the interactions of multiple TFs and RBPs with nucleic acids to identify their common and unique targets is necessary. Therefore, we introduce BindCompare, a user-friendly tool that can be run locally to predict new combinatorial relationships between TFs and RBPs. BindCompare can analyze data from any organism with known annotated genome information and outputs files with detailed genomic locations and gene information for targets for downstream analysis. Overall, BindCompare is a new tool that identifies TFs and RBPs that co-bind to the same DNA and/or RNA loci, generating testable hypotheses about their combinatorial regulation of target genes. BindCompare is an open-source package that is available on the Python Packaging Index (PyPI, https://pypi.org/project/bindcompare/) with the source code available on GitHub (https://github.com/pranavmahabs/bindcompare). Complete documentation for the package can be found at both links. Supplementary data are available at Bioinformatics online.

Network dynamics-based subtyping of Alzheimer’s disease with microglial genetic risk factors

BackgroundThe potential of microglia as a target for Alzheimer’s disease (AD) treatment is promising, yet the clinical and pathological diversity within microglia, driven by genetic factors, poses a significant challenge. Subtyping AD is imperative to enable precise and effective treatment strategies. However, existing subtyping methods fail to comprehensively address the intricate complexities of AD pathogenesis, particularly concerning genetic risk factors. To address this gap, we have employed systems biology approaches for AD subtyping and identified potential therapeutic targets.MethodsWe constructed patient-specific microglial molecular regulatory network models by utilizing existing literature and single-cell RNA sequencing data. The combination of large-scale computer simulations and dynamic network analysis enabled us to subtype AD patients according to their distinct molecular regulatory mechanisms. For each identified subtype, we suggested optimal targets for effective AD treatment.ResultsTo investigate heterogeneity in AD and identify potential therapeutic targets, we constructed a microglia molecular regulatory network model. The network model incorporated 20 known risk factors and crucial signaling pathways associated with microglial functionality, such as inflammation, anti-inflammation, phagocytosis, and autophagy. Probabilistic simulations with patient-specific genomic data and subsequent dynamics analysis revealed nine distinct AD subtypes characterized by core feedback mechanisms involving SPI1, CASS4, and MEF2C. Moreover, we identified PICALM, MEF2C, and LAT2 as common therapeutic targets among several subtypes. Furthermore, we clarified the reasons for the previous contradictory experimental results that suggested both the activation and inhibition of AKT or INPP5D could activate AD through dynamic analysis. This highlights the multifaceted nature of microglial network regulation.ConclusionsThese results offer a means to classify AD patients by their genetic risk factors, clarify inconsistent experimental findings, and advance the development of treatments tailored to individual genotypes for AD.

MicroRNAs provide negative feedback and stability in gene regulatory network models of cell-state transitions

The development of multicellular organisms occurs through a series of cell state transitions controlled by gene regulatory networks. Central to these networks are transcription factors (TFs) which bind enhancers and activate the expression of other genes, some of which are also TFs. Gene regulatory networks (GRN) connect TFs and enhancers in a nonlinear circuit capable of producing complex behavior such as bifurcations between stable cell states. Our dynamic network modelling of the Embryonic Stem Cell (ESC) to Definitive Endoderm (DE) transition requires an as yet unknown negative feedback mechanism for stability. Here, we show that cell state specific microRNAs (miRNAs) can provide this negative feedback by inactivating other cell lineage determining TFs (ESC or DE) during the transition. Our model provides a mechanism to maintain stable cell states without requiring a large set of cell-type-specific repressive TFs, of which there are fewer known examples than activators. In support of this model, we use computational models and analyze gene and miRNA expression and chromatin accessibility data from human cell lines to detect enhancers activating the miRNAs consistent with our network model. Our analysis highlights the interplay between TFs and miRNAs during ESC to DE transition and proposes a novel model for gene regulation.

Merging logical models: An application in Acute Myeloid Leukemia modeling.

Gene regulatory network (GRNs) models provide mechanistic understanding of the gene regulations and interactions that control various aspects of cellular behaviors. While researchers have constructed GRNs to model specific sets of gene regulations or interactions, little work has been made to integrate or merge these models into larger, more comprehensive ones that could encompass more genes, and improve the accuracy of predicting biological processes. Here, we present a workflow for merging logical GRN models, which requires sequential steps including model standardization, reproducing, merging and evaluations, and demonstrate its application in acute myeloid leukemia (AML) study. We demonstrate the feasibility and benefits of model merging by integrating two pairs of published models. Our integrated models were able to retain similar accuracy of the original publications, while increasing the coverage and explainability of the biological system. This approach highlights the integration of logical models in advancing system biology and enhancing the understanding of complex diseases.

Tipping-point transition from transient to persistent inflammation in pancreatic islets

Type 2 diabetes (T2D) is associated with a systemic increase in the pro-inflammatory cytokine IL-1β. While transient exposure to low IL-1β concentrations improves insulin secretion and β-cell proliferation in pancreatic islets, prolonged exposure leads to impaired insulin secretion and collective β-cell death. IL-1 is secreted locally by islet-resident macrophages and β-cells; however, it is unknown if and how the two opposing modes may emerge at single islet level. We investigated the duality of IL-1β with a quantitative in silico model of the IL-1 regulatory network in pancreatic islets. We find that the network can produce either transient or persistent IL-1 responses when induced by pro-inflammatory and metabolic cues. This suggests that the duality of IL-1 may be regulated at the single islet level. We use two core feedbacks in the IL-1 regulation to explain both modes: First, a fast positive feedback in which IL-1 induces its own production through the IL-1R/IKK/NF-κB pathway. Second, a slow negative feedback where NF-κB upregulates inhibitors acting at different levels along the IL-1R/IKK/NF-κB pathway—IL-1 receptor antagonist and A20, among others. A transient response ensues when the two feedbacks are balanced. When the positive feedback dominates over the negative, islets transit into the persistent inflammation mode. Consistent with several observations, where the size of islets was implicated in its inflammatory state, we find that large islets and islets with high density of IL-1β amplifying cells are more prone to transit into persistent IL-1β mode. Our results are likely not limited to IL-1β but are general for the combined effect of multiple pro-inflammatory cytokines and chemokines. Generalizing complex regulations in terms of two feedback mechanisms of opposing nature and acting on different time scales provides a number of testable predictions. Taking islet architecture and cellular heterogeneity into consideration, further dynamic monitoring and experimental validation in actual islet samples will be crucial to verify the model predictions and enhance its utility in clinical applications.

Decoding the plant clock: a review of mathematical models for the circadian regulatory network.

Most organisms have evolved specific mechanisms to respond to changes in environmental conditions such as light and temperature over the course of day. These periodic changes in the physiology and behaviour of organisms, referred to as circadian rhythms, are a consequence of intricate molecular mechanisms in the form of transcription and translational feedback loops. The plant circadian regulatory network is a complex web of interconnected feedback loops involving various transcription factors such as CCA1, LHY, PRRs, TOC1, LUX, ELF3, ELF4, RVE8, and more. This network enables plants to adapt and thrive in diverse environmental conditions. It responds to entrainment signals, including light, temperature, and nutrient concentrations and interacts with most of the physiological functions such as flowering, growth and stress response. Mathematical modelling of these gene regulatory networks enables a deeper understanding of not only the function but also the perturbations that may affect the plant growth and function with changing climate. Over the years, numerous mathematical models have been developed to understand the diverse aspects of plant circadian regulation. In this review, we have delved into the systematic development of these models, outlining the model components and refinements over time. We have also highlighted strengths and limitations of each of the models developed so far. Finally, we conclude the review by describing the prospects for investigation and advancement of these models for better understanding of plant circadian regulation.

Unlocking mitochondrial dysfunction-associated senescence (MiDAS) with NAD+ – A Boolean model of mitochondrial dynamics and cell cycle control

The steady accumulation of senescent cells with aging creates tissue environments that aid cancer evolution. Aging cell states are highly heterogeneous. 'Deep senescent' cells rely on healthy mitochondria to fuel a strong proinflammatory secretome, including cytokines, growth and transforming signals. Yet, the physiological triggers of senescence such as reactive oxygen species (ROS) can also trigger mitochondrial dysfunction, and sufficient energy deficit to alter their secretome and cause chronic oxidative stress – a state termed Mitochondrial Dysfunction-Associated Senescence (MiDAS). Here, we offer a mechanistic hypothesis for the molecular processes leading to MiDAS, along with testable predictions. To do this we have built a Boolean regulatory network model that qualitatively captures key aspects of mitochondrial dynamics during cell cycle progression (hyper-fusion at the G1/S boundary, fission in mitosis), apoptosis (fission and dysfunction) and glucose starvation (reversible hyper-fusion), as well as MiDAS in response to SIRT3 knockdown or oxidative stress. Our model reaffirms the protective role of NAD+ and external pyruvate. We offer testable predictions about the growth factor- and glucose-dependence of MiDAS and its reversibility at different stages of reactive oxygen species (ROS)-induced senescence. Our model provides mechanistic insights into the distinct stages of DNA-damage induced senescence, the relationship between senescence and epithelial-to-mesenchymal transition in cancer and offers a foundation for building multiscale models of tissue aging.

Transcriptome Profiling of Mouse Embryonic Fibroblast Spontaneous Immortalization: A Comparative Analysis.

Cell immortalization, a hallmark of cancer development, is a process that cells can undergo on their path to carcinogenesis. Spontaneously immortalized mouse embryonic fibroblasts (MEFs) have been used for decades; however, changes in the global transcriptome during this process have been poorly described. In our research, we characterized the poly-A RNA transcriptome changes after spontaneous immortalization. To this end, differentially expressed genes (DEGs) were screened using DESeq2 and characterized by gene ontology enrichment analysis and protein-protein interaction (PPI) network analysis to identify the potential hub genes. In our study, we identified changes in the expression of genes involved in proliferation regulation, cell adhesion, immune response and transcriptional regulation in immortalized MEFs. In addition, we performed a comparative analysis with previously reported MEF immortalization data, where we propose a predicted gene regulatory network model in immortalized MEFs based on the altered expression of Mapk11, Cdh1, Chl1, Zic1, Hoxd10 and the novel hub genes Il6 and Itgb2.

On the evolutionary origin of discrete phenotypic plasticity.

Phenotypic plasticity provides an attractive strategy for adapting to various environments, but the evolutionary mechanism of the underlying genetic system is poorly understood. We use a simple gene regulatory network model to explore how a species acquires phenotypic plasticity, particularly focusing on discrete phenotypic plasticity, which has been difficult to explain by quantitative genetic models. Our approach employs a population genetic framework that integrates the developmental process, where each individual undergoes growth to develop its phenotype, which subsequently becomes subject to selection pressures. Our model considers two alternative types of environments, with the gene regulatory network including a sensor gene that turns on and off depending on the type of environment. With this assumption, we demonstrate that the system gradually adapts by acquiring the ability to produce two distinct optimum phenotypes under two types of environments without changing genotype, resulting in phenotypic plasticity. We find that the resulting plasticity is often discrete after a lengthy period of evolution. Our results suggest that gene regulatory networks have a notable capacity to flexibly produce various phenotypes in response to environmental changes. This study also shows that the evolutionary dynamics of phenotype may differ significantly between mechanistic-based developmental models and quantitative genetics models, suggesting the utility of incorporating gene regulatory networks into evolutionary models.

Characterizing behavioural differentiation in gene regulatory networks with representation graphs.

We introduce the formal notion of representation graphs, encapsulating the state space structure of gene regulatory network models in a compact and concise form that highlights the most significant features of stable states and differentiation processes leading to distinct stability regions. The concept has been developed in the context of a hybrid system-based gene network modelling framework; however, we anticipate that it can also be adapted to other approaches of modelling gene networks in discrete terms. We describe a practical algorithm for representation graph computation as well as two case studies demonstrating their real-world application and utility. The first case study presents models for three phage viruses. It shows that the process of differentiation into lytic and lysogenic behavioural states for all these models is described by the same representation graph despite the distinctive underlying mechanisms for differentiation. The second case study shows the advantages of our approach for modelling the process of myeloid cell differentiation from a common progenitor into different cell types. Both case studies also demonstrate the potential of the representation graph approach for deriving and validating hypotheses about regulatory interactions that must be satisfied for biologically viable behaviours.

Data-driven mathematical modeling of sleep consolidation in early childhood

Across early childhood development, sleep behavior transitions from a biphasic pattern (a daytime nap and nighttime sleep) to a monophasic pattern (only nighttime sleep). The transition to consolidated nighttime sleep, which occurs in most children between 2- and 5-years-old, is a major developmental milestone and reflects interactions between the developing homeostatic sleep drive and circadian system. Using a physiologically-based mathematical model of the sleep-wake regulatory network constrained by observational and experimental data from preschool-aged participants, we analyze how developmentally-mediated changes in the homeostatic sleep drive may contribute to the transition from napping to non-napping sleep patterns. We establish baseline behavior by identifying parameter sets that model typical 2-year-old napping behavior and 5-year-old non-napping behavior. Then we vary six model parameters associated with the dynamics of and sensitivity to the homeostatic sleep drive between the 2-year-old and 5-year-old parameter values to induce the transition from biphasic to monophasic sleep. We analyze the individual contributions of these parameters to sleep patterning by independently varying their age-dependent developmental trajectories. Parameters vary according to distinct evolution curves and produce bifurcation sequences representing various ages of transition onset, transition durations, and transitional sleep patterns. Finally, we consider the ability of napping and non-napping light schedules to reinforce napping or promote a transition to consolidated sleep, respectively. These modeling results provide insight into the role of the homeostatic sleep drive in promoting interindividual variability in developmentally-mediated transitions in sleep behavior and lay foundations for the identification of light- or behavior-based interventions that promote healthy sleep consolidation in early childhood.

SAILoR: Structure-Aware Inference of Logic Rules.

Boolean networks provide an effective mechanism for describing interactions and dynamics of gene regulatory networks (GRNs). Deriving accurate Boolean descriptions of GRNs is a challenging task. The number of experiments is usually much smaller than the number of genes. In addition, binarization leads to a loss of information and inconsistencies arise in binarized time-series data. The inference of Boolean networks from binarized time-series data alone often leads to complex and overfitted models. To obtain relevant Boolean models of gene regulatory networks, inference methods could incorporate data from multiple sources and prior knowledge in terms of general network structure and/or exact interactions. We propose the Boolean network inference method SAILoR (Structure-Aware Inference of Logic Rules). SAILoR incorporates time-series gene expression data in combination with provided reference networks to infer accurate Boolean models. SAILoR automatically extracts topological properties from reference networks. These can describe a more general structure of the GRN or can be more precise and describe specific interactions. SAILoR infers a Boolean network by learning from both continuous and binarized time-series data. It navigates between two main objectives, topological similarity to reference networks and correspondence with gene expression data. By incorporating the NSGA-II multi-objective genetic algorithm, SAILoR relies on the wisdom of crowds. Our results indicate that SAILoR can infer accurate and biologically relevant Boolean descriptions of GRNs from both a static and a dynamic perspective. We show that SAILoR improves the static accuracy of the inferred network compared to the network inference method dynGENIE3. Furthermore, we compared the performance of SAILoR with other Boolean network inference approaches including Best-Fit, REVEAL, MIBNI, GABNI, ATEN, and LogBTF. We have shown that by incorporating prior knowledge about the overall network structure, SAILoR can improve the structural correctness of the inferred Boolean networks while maintaining dynamic accuracy. To demonstrate the applicability of SAILoR, we inferred context-specific Boolean subnetworks of female Drosophila melanogaster before and after mating.

Uniform convergence of solutions to stochastic hybrid models of gene regulatory networks

In a recent paper by Kurasov et al. (Math Biosci 305:170–177, 2018), a hybrid gene regulatory network was proposed to model gene expression dynamics by using a stochastic system of coupled partial differential equations. In more recent work, the existence and strong convergence of the solutions to equilibrium were proven. In this article, we improve upon their result by showing that the convergence rate is independent of the initial state, therefore proving that the solutions converge not only strongly but even uniformly to equilibrium. To this end, we make use of a recent convergence theorem for stochastic, irreducible semigroups that contain partial integral operators.

D-SPIN constructs gene regulatory network models from multiplexed scRNA-seq data revealing organizing principles of cellular perturbation response.

Gene regulatory networks within cells modulate the expression of the genome in response to signals and changing environmental conditions. Reconstructions of gene regulatory networks can reveal the information processing and control principles used by cells to maintain homeostasis and execute cell-state transitions. Here, we introduce a computational framework, D-SPIN, that generates quantitative models of gene regulatory networks from single-cell mRNA-seq datasets collected across thousands of distinct perturbation conditions. D-SPIN models the cell as a collection of interacting gene-expression programs, and constructs a probabilistic model to infer regulatory interactions between gene-expression programs and external perturbations. Using large Perturb-seq and drug-response datasets, we demonstrate that D-SPIN models reveal the organization of cellular pathways, sub-functions of macromolecular complexes, and the logic of cellular regulation of transcription, translation, metabolism, and protein degradation in response to gene knockdown perturbations. D-SPIN can also be applied to dissect drug response mechanisms in heterogeneous cell populations, elucidating how combinations of immunomodulatory drugs can induce novel cell states through additive recruitment of gene expression programs. D-SPIN provides a computational framework for constructing interpretable models of gene-regulatory networks to reveal principles of cellular information processing and physiological control.

ATP dependence of decision-making capacity in a fine-grained model of gene regulatory networks

Cellular decision-making is fundamental to life, from developmental biology to environmental responses and antimicrobial resistance. Many regulatory processes that drive cellular decisions rely on gene expression, which requires energy in the form of ATP. As even genetically identical cells can have dramatically different ATP levels, bioenergetic status can be an important source of variability in cellular decision-making. Existing studies have investigated this energy dependence but often use coarse-grained modelling approaches (which are not always readily connected to the underlying molecular processes of gene regulation). Here, we use a fine-grained mathematical model of gene expression in a two-gene decision-making regulatory network to explore cellular decision-making capacity as energy availability varies. We simulate both a deterministic model, to explore the emergence of different cell fate attractors as ATP levels vary, and a stochastic case to explore how ATP influences the noisy dynamics of stochastic cell decision-making. Higher energy levels typically support increased decision-making capacity (higher numbers of, and more separated, cell states that can be selected), and the fine-grained modelling reveals some differences in behaviour from previous coarse-grained modelling approaches.

Identification of Potential lncRNA-miRNA-mRNA Regulatory Network Contributing to Arrhythmogenic Right Ventricular Cardiomyopathy.

Arrhythmogenic right ventricular cardiomyopathy (ARVC) can lead to sudden cardiac death and life-threatening heart failure. Due to its high fatality rate and limited therapies, the pathogenesis and diagnosis biomarker of ARVC needs to be explored urgently. This study aimed to explore the lncRNA-miRNA-mRNA competitive endogenous RNA (ceRNA) network in ARVC. The mRNA and lncRNA expression datasets obtained from the Gene Expression Omnibus (GEO) database were used to analyze differentially expressed mRNA (DEM) and lncRNA (DElnc) between ARVC and non-failing controls. Differentially expressed miRNAs (DEmiRs) were obtained from the previous profiling work. Using starBase to predict targets of DEmiRs and intersecting with DEM and DElnc, a ceRNA network of lncRNA-miRNA-mRNA was constructed. The DEM and DElnc were validated by real-time quantitative PCR in human heart tissue. Protein-protein interaction network and weighted gene co-expression network analyses were used to identify hub genes. A logistic regression model for ARVC diagnostic prediction was established with the hub genes and their ceRNA pairs in the network. A total of 448 DEMs (282 upregulated and 166 downregulated) were identified, mainly enriched in extracellular matrix and fibrosis-related GO terms and KEGG pathways, such as extracellular matrix organization and collagen fibril organization. Four mRNAs and two lncRNAs, including COL1A1, COL5A1, FBN1, BGN, XIST, and LINC00173 identified through the ceRNA network, were validated by real-time quantitative PCR in human heart tissue and used to construct a logistic regression model. Good ARVC diagnostic prediction performance for the model was shown in both the training set and the validation set. The potential lncRNA-miRNA-mRNA regulatory network and logistic regression model established in our study may provide promising diagnostic methods for ARVC.

Flexible modeling of regulatory networks improves transcription factor activity estimation

Transcriptional regulation plays a crucial role in determining cell fate and disease, yet inferring the key regulators from gene expression data remains a significant challenge. Existing methods for estimating transcription factor (TF) activity often rely on static TF-gene interaction databases and cannot adapt to changes in regulatory mechanisms across different cell types and disease conditions. Here, we present a new algorithm - Transcriptional Inference using Gene Expression and Regulatory data (TIGER) - that overcomes these limitations by flexibly modeling activation and inhibition events, up-weighting essential edges, shrinking irrelevant edges towards zero through a sparse Bayesian prior, and simultaneously estimating both TF activity levels and changes in the underlying regulatory network. When applied to yeast and cancer TF knock-out datasets, TIGER outperforms comparable methods in terms of prediction accuracy. Moreover, our application of TIGER to tissue- and cell-type-specific RNA-seq data demonstrates its ability to uncover differences in regulatory mechanisms. Collectively, our findings highlight the utility of modeling context-specific regulation when inferring transcription factor activities.

Genome-wide analysis of long noncoding RNAs as cis-acting regulators of transcription factor-encoding genes in IgA nephropathy.

IgA nephropathy (IgAN) is the most common form of primary glomerulonephritis in the world, but the disease pathogenesis noncoding is yet to be elucidated. Previous studies have revealed regulatory functions for long noncoding RNA (lncRNA) in various diseases; however, the roles of lncRNA in IgAN and regulation of transcription factors (TFs) have been scarcely investigated. Renal tissue samples (n = 5) from patients with IgAN and control samples (n = 4) were collected and RNA sequencing (RNA-seq) was performed. Four software programs were employed for lncRNA prediction. GO (Gene Ontology)/KEGG (Kyoto Encyclopedia of Genes and Genomes) were employed for analysis of the identified differentially expressed genes (DEGs). A regulatory network model of DE lncRNA-TF-DEG was developed, and the levels of expression of key lncRNAs, TFs, and corresponding target genes were assessed using qRT-PCR and immunofluorescence. The current study identified 674 upregulated and 1,011 downregulated DE mRNAs and 260 upregulated and 232 downregulated DE lncRNAs in IgAN samples compared with control samples. The upregulated DE mRNAs showed enrichment in cell adhesion and collagen glial fiber organization pathways. The DE lncRNAs-DE mRNAs showing co-expression are associated with transmembrane transport. A novel regulatory network model of lncRNA-TF-DEG has been developed. This study identified seven TFs that are cis-regulated by 6 DE lncRNAs, and show co-expression with 132 DEGs (correlation coefficient ≥ 0.8, P ≤ 0.01), generating 158 pairs that showed co-expression. The lncRNAs NQO1-DT and RP5-1057120.6 were found to be highly expressed in IgAN samples. The TFs vitamin D Receptor (VDR) and NFAT5, along with their target genes were also aberrantly expressed. Key lncRNAs and TFs centrally associated with IgAN have been identified in this study. A regulatory network model of lncRNA-TF-mRNA was constructed. Further studies on the genes identified herewith could provide insight into the pathogenesis of IgAN.

Characterization of clinical data for patient stratification in moderate osteoarthritis with support vector machines, regulatory network models, and verification against osteoarthritis Initiative data

Knee osteoarthritis (OA) diagnosis is based on symptoms, assessed through questionnaires such as the WOMAC. However, the inconsistency of pain recording and the discrepancy between joint phenotype and symptoms highlight the need for objective biomarkers in knee OA diagnosis. To this end, we study relationships among clinical and molecular data in a cohort of women (n = 51) with Kellgren–Lawrence grade 2–3 knee OA through a Support Vector Machine (SVM) and a regulation network model. Clinical descriptors (i.e., pain catastrophism, depression, functionality, joint pain, rigidity, sensitization and synovitis) are used to classify patients. A Youden’s test is performed for each classifier to determine optimal binarization thresholds for the descriptors. Thresholds are tested against patient stratification according to baseline WOMAC data from the Osteoarthritis Initiative, and the mean accuracy is 0.97. For our cohort, the data used as SVM inputs are knee OA descriptors, synovial fluid proteomic measurements (n = 25), and transcription factor activation obtained from regulatory network model stimulated with the synovial fluid measurements. The relative weights after classification reflect input importance. The performance of each classifier is evaluated through ROC-AUC analysis. The best classifier with clinical data is pain catastrophism (AUC = 0.9), highly influenced by funcionality and pain sensetization, suggesting that kinesophobia is involved in pain perception. With synovial fluid proteins used as input, leptin strongly influences every classifier, suggesting the importance of low-grade inflammation. When transcription factors are used, the mean AUC is limited to 0.608, which can be related to the pleomorphic behaviour of osteoarthritic chondrocytes. Nevertheless, funcionality has an AUC of 0.7 with a decisive importance of FOXO downregulation. Though larger and longitudinal cohorts are needed, this unique combination of SVM and regulatory network model shall help to stratify knee OA patients more objectively.

Biologically informed NeuralODEs for genome-wide regulatory dynamics

BackgroundGene regulatory network (GRN) models that are formulated as ordinary differential equations (ODEs) can accurately explain temporal gene expression patterns and promise to yield new insights into important cellular processes, disease progression, and intervention design. Learning such gene regulatory ODEs is challenging, since we want to predict the evolution of gene expression in a way that accurately encodes the underlying GRN governing the dynamics and the nonlinear functional relationships between genes. Most widely used ODE estimation methods either impose too many parametric restrictions or are not guided by meaningful biological insights, both of which impede either scalability, explainability, or both.ResultsWe developed PHOENIX, a modeling framework based on neural ordinary differential equations (NeuralODEs) and Hill-Langmuir kinetics, that overcomes limitations of other methods by flexibly incorporating prior domain knowledge and biological constraints to promote sparse, biologically interpretable representations of GRN ODEs. We tested the accuracy of PHOENIX in a series of in silico experiments, benchmarking it against several currently used tools. We demonstrated PHOENIX’s flexibility by modeling regulation of oscillating expression profiles obtained from synchronized yeast cells. We also assessed the scalability of PHOENIX by modeling genome-scale GRNs for breast cancer samples ordered in pseudotime and for B cells treated with Rituximab.ConclusionsPHOENIX uses a combination of user-defined prior knowledge and functional forms from systems biology to encode biological “first principles” as soft constraints on the GRN allowing us to predict subsequent gene expression patterns in a biologically explainable manner.