Search

Large-scale studies of gene expression are commonly influenced by biological and technical sources of expression variation, including batch effects, sample characteristics, and environmental impacts. Learning the causal relationships between observable variables may be challenging in the presence of unobserved confounders. Furthermore, many high-dimensional regression techniques may perform worse. In fact, controlling for unobserved confounding variables is essential, and many deconfounding methods have been suggested for application in a variety of situations. The main contribution of this article is the development of a two-stage deconfounding procedure based on Bow-free Acyclic Paths (BAP) search developed into the framework of Structural Equation Models (SEM), called SEMbap(). In the first stage, an exhaustive search of missing edges with significant covariance is performed via Shipley d-separation tests; then, in the second stage, a Constrained Gaussian Graphical Model (CGGM) is fitted or a low dimensional representation of bow-free edges structure is obtained via Graph Laplacian Principal Component Analysis (gLPCA). We compare four popular deconfounding methods to BAP search approach with applications on simulated and observed expression data. In the former, different structures of the hidden covariance matrix have been replicated. Compared to existing methods, BAP search algorithm is able to correctly identify hidden confounding whilst controlling false positive rate and achieving good fitting and perturbation metrics. Author summary: Directed acyclic graphs (DAGs) directed graph, with variables at the vertices and direct causal connections at the edges, can be used to illustrate the causal structure of the SEM, but this does not always mean that all significant factors are considered. We examine a class of models that may include some hidden variables. Specifically, we consider that the graph represents a bow-free acyclic path diagram (BAP), where the directed edges signify direct causal effects, while the bidirected edges suggest hidden confounders. In this paper, we provide a two-step deconfounding technique based on BAP search, which is included into the SEM framework via the SEMbap() function implemented in the R package SEMgraph. Secondly, we want to offer a significant evaluation of the most advanced deconfounding techniques using both synthetic and real data, as well as knowledge of a biological signaling pathway encoded in a DAG, in terms of (i) SEM fitting, (ii) system perturbation, and (iii) recovery performance metrics. The BAP search algorithm outperforms current techniques in accurately detecting hidden confounding, regulating false positive rate, and producing well-fitting and perturbation metrics. [ABSTRACT FROM AUTHOR]

Boolean networks are largely employed to model the qualitative dynamics of cell fate processes by describing the change of binary activation states of genes and transcription factors with time. Being able to bridge such qualitative states with quantitative measurements of gene expressions in cells, as scRNA-seq, is a cornerstone for data-driven model construction and validation. On one hand, scRNA-seq binarisation is a key step for inferring and validating Boolean models. On the other hand, the generation of synthetic scRNA-seq data from baseline Boolean models provides an important asset to benchmark inference methods. However, linking characteristics of scRNA-seq datasets, including dropout events, with Boolean states is a challenging task. We present scBoolSeq, a method for the bidirectional linking of scRNA-seq data and Boolean activation state of genes. Given a reference scRNA-seq dataset, scBoolSeq computes statistical criteria to classify the empirical gene pseudocount distributions as either unimodal, bimodal, or zero-inflated, and fit a probabilistic model of dropouts, with gene-dependent parameters. From these learnt distributions, scBoolSeq can perform both binarisation of scRNA-seq datasets, and generate synthetic scRNA-seq datasets from Boolean traces, as issued from Boolean networks, using biased sampling and dropout simulation. We present a case study demonstrating the application of scBoolSeq's binarisation scheme in data-driven model inference. Furthermore, we compare synthetic scRNA-seq data generated by scBoolSeq with BoolODE's, data for the same Boolean Network model. The comparison shows that our method better reproduces the statistics of real scRNA-seq datasets, such as the mean-variance and mean-dropout relationships while exhibiting clearly defined trajectories in two-dimensional projections of the data. Author summary: The qualitative and logical modelling of cell dynamics has brought precious insight into gene regulatory mechanisms that drive cellular differentiation and fate decisions by predicting cellular trajectories and mutations for their control. However, the design and validation of these models is impeded by the quantitative nature of experimental measurements of cellular states. In this paper, we provide and assess a new methodology, scBoolSeq for bridging single-cell level pseudocounts of RNA transcripts with Boolean classification of gene activity levels. Our method, implemented as a Python package, enables both to binarise scRNA-seq data in order to match quantitative measurements with states of logical models, and to generate synthetic data from Boolean traces to benchmark inference methods. We show that scBoolSeq accurately captures the main statistical features of scRNA-seq data, including measurement dropouts, improving significantly the state of the art. Overall, scBoolSeq brings a statistically-grounded method for enabling the inference and validation of qualitative models from scRNA-seq data. [ABSTRACT FROM AUTHOR]

In experiments, the distributions of mRNA or protein numbers in single cells are often fitted to the random telegraph model which includes synthesis and decay of mRNA or protein, and switching of the gene between active and inactive states. While commonly used, this model does not describe how fluctuations are influenced by crucial biological mechanisms such as feedback regulation, non-exponential gene inactivation durations, and multiple gene activation pathways. Here we investigate the dynamical properties of four relatively complex gene expression models by fitting their steady-state mRNA or protein number distributions to the simple telegraph model. We show that despite the underlying complex biological mechanisms, the telegraph model with three effective parameters can accurately capture the steady-state gene product distributions, as well as the conditional distributions in the active gene state, of the complex models. Some effective parameters are reliable and can reflect realistic dynamic behaviors of the complex models, while others may deviate significantly from their real values in the complex models. The effective parameters can also be applied to characterize the capability for a complex model to exhibit multimodality. Using additional information such as single-cell data at multiple time points, we provide an effective method of distinguishing the complex models from the telegraph model. Furthermore, using measurements under varying experimental conditions, we show that fitting the mRNA or protein number distributions to the telegraph model may even reveal the underlying gene regulation mechanisms of the complex models. The effectiveness of these methods is confirmed by analysis of single-cell data for E. coli and mammalian cells. All these results are robust with respect to cooperative transcriptional regulation and extrinsic noise. In particular, we find that faster relaxation speed to the steady state results in more precise parameter inference under large extrinsic noise. Author summary: Over the past decade, significant progress has been made in the theory and experiments of single-cell stochastic gene expression dynamics. The most well studied and widely used stochastic gene expression model is the two-state telegraph model. However, the conventional telegraph model is too simple and limited in its predictive power because it lacks a description of some important biological mechanisms, such as feedback regulation, multiple gene activation steps, and multiple gene activation pathways. This raises a important question: what can we learn when fitting a complex gene expression model to a simple telegraph model? In this paper, we investigate four complex gene expression models by fitting their steady-state mRNA or protein number distributions to the telegraph model and then obtain estimates of the effective parameters. We show that while the estimated values of the parameters in the "artificial" telegraph model are not always accurate, they are still sometimes reliable and can also reveal important dynamical properties of the complex models such as the ability for a complex model to produce bimodality. Moreover, we provide an effective method of distinguishing the complex models from the telegraph model by using additional information such as gene expression data at multiple time points. Finally, we show that fitting the mRNA or protein number distributions to the telegraph model may even reveal the underlying gene regulation mechanism of a complex model by using measurements under varying experimental conditions. The effectiveness of these methods is well confirmed by analysis of single-cell gene expression data for E. coli and mammalian cells. [ABSTRACT FROM AUTHOR]

The heterogeneity of Hepatocellular Carcinoma (HCC) poses a barrier to effective treatment. Stratifying highly heterogeneous HCC into molecular subtypes with similar features is crucial for personalized anti-tumor therapies. Although driver genes play pivotal roles in cancer progression, their potential in HCC subtyping has been largely overlooked. This study aims to utilize driver genes to construct HCC subtype models and unravel their molecular mechanisms. Utilizing a novel computational framework, we expanded the initially identified 96 driver genes to 1192 based on mutational aspects and an additional 233 considering driver dysregulation. These genes were subsequently employed as stratification markers for further analyses. A novel multi-omics subtype classification algorithm was developed, leveraging mutation and expression data of the identified stratification genes. This algorithm successfully categorized HCC into two distinct subtypes, CLASS A and CLASS B, demonstrating significant differences in survival outcomes. Integrating multi-omics and single-cell data unveiled substantial distinctions between these subtypes regarding transcriptomics, mutations, copy number variations, and epigenomics. Moreover, our prognostic model exhibited excellent predictive performance in training and external validation cohorts. Finally, a 10-gene classification model for these subtypes identified TTK as a promising therapeutic target with robust classification capabilities. This comprehensive study provides a novel perspective on HCC stratification, offering crucial insights for a deeper understanding of its pathogenesis and the development of promising treatment strategies. Author summary: Dividing highly heterogeneous HCC into molecular subtypes with similar characteristics is crucial for personalized anti-tumor therapies. Although driver genes play pivotal roles in cancer progression, their potential in HCC subtyping has been largely overlooked. In this work, we developed a multi-omics network-based stratification algorithm that utilizes patient mutation data and requires smaller computational resources for subtype assignment. Through this algorithm, we categorized HCC into two subtypes, CLASS A and CLASS B. Using multi-omics and single-cell data, we identified differences between these subtypes in gene expression, methylation, immune infiltration, and other aspects. Beyond subtype characterization, our study established a robust clinical prediction model (https://mike-wang-bjut.shinyapps.io/DynNomapp%5fHCC%5fSutypes/) incorporating subtype information and typical clinical features, enabling precise survival predictions. Finally, we developed a high-performing machine learning classifier for our subtype. Analyzing this classification model and reviewing previous experimental papers, we identified TTK as a potential diagnostic marker and therapeutic target specific to our subtypes. In conclusion, our research offers a novel perspective on HCC stratification, which is crucial for a deeper understanding of its pathogenesis and developing promising treatment strategies. [ABSTRACT FROM AUTHOR]

A core problem in genetics is molecular quantitative trait locus (QTL) mapping, in which genetic variants associated with changes in the molecular phenotypes are identified. One of the most-studied molecular QTL mapping problems is expression QTL (eQTL) mapping, in which the molecular phenotype is gene expression. It is common in eQTL mapping to compute gene expression by aggregating the expression levels of individual isoforms from the same gene and then performing linear regression between SNPs and this aggregated gene expression level. However, SNPs may regulate isoforms from the same gene in different directions due to alternative splicing, or only regulate the expression level of one isoform, causing this approach to lose power. Here, we examine a broader question: which genes have at least one isoform whose expression level is regulated by genetic variants? In this study, we propose and evaluate several approaches to answering this question, demonstrating that "isoform-aware" methods—those that account for the expression levels of individual isoforms—have substantially greater power to answer this question than standard "gene-level" eQTL mapping methods. We identify settings in which different approaches yield an inflated number of false discoveries or lose power. In particular, we show that calling an eGene if there is a significant association between a SNP and any isoform fails to control False Discovery Rate, even when applying standard False Discovery Rate correction. We show that similar trends are observed in real data from the GEUVADIS and GTEx studies, suggesting the possibility that similar effects are present in these consortia. [ABSTRACT FROM AUTHOR]

We describe a novel single sample gene set testing method for cancer transcriptomics data named tissue-adjusted pathway analysis of cancer (TPAC). The TPAC method leverages information about the normal tissue-specificity of human genes to compute a robust multivariate distance score that quantifies gene set dysregulation in each profiled tumor. Because the null distribution of the TPAC scores has an accurate gamma approximation, both population and sample-level inference is supported. As we demonstrate through an analysis of gene expression data for 21 solid human cancers from The Cancer Genome Atlas (TCGA) and associated normal tissue expression data from the Human Protein Atlas (HPA), TPAC gene set scores are more strongly associated with patient prognosis than the scores generated by existing single sample gene set testing methods. Author summary: Cancer biology is highly tissue-specific: most cancer-driving somatic alterations occur in just a limited number of tissues and inherited mutations frequently have a tissue-specific functional impact. To leverage the associations between gene activity in normal and malignant tissue for gene set testing, we have developed a new single sample gene set testing method for tumor-derived transcriptomics data named TPAC (tissue-adjusted pathway analysis of cancer). The TPAC method uses normal tissue-specificity to quantifies gene set dysregulation in each profiled tumor. Importantly, we found that TPAC gene set scores are more strongly associated with patient prognosis than the scores generated by existing single sample gene set testing methods. [ABSTRACT FROM AUTHOR]

Inferring the driving regulatory programs from comparative analysis of gene expression data is a cornerstone of systems biology. Many computational frameworks were developed to address this problem, including our iPAGE (information-theoretic Pathway Analysis of Gene Expression) toolset that uses information theory to detect non-random patterns of expression associated with given pathways or regulons. Our recent observations, however, indicate that existing approaches are susceptible to the technical biases that are inherent to most real world annotations. To address this, we have extended our information-theoretic framework to account for specific biases and artifacts in biological networks using the concept of conditional information. To showcase pyPAGE, we performed a comprehensive analysis of regulatory perturbations that underlie the molecular etiology of Alzheimer's disease (AD). pyPAGE successfully recapitulated several known AD-associated gene expression programs. We also discovered several additional regulons whose differential activity is significantly associated with AD. We further explored how these regulators relate to pathological processes in AD through cell-type specific analysis of single cell and spatial gene expression datasets. Our findings showcase the utility of pyPAGE as a precise and reliable biomarker discovery in complex diseases such as Alzheimer's disease. Author summary: Biological regulation is governed by a complex network of interactions involving transcription factors, RNA-binding proteins, and microRNAs. To reveal the regulatory programs underlying gene expression modulations, researchers often take advantage of gene-set enrichment analysis, an approach that studies concerted changes in a group of genes rather than observing genes in isolation. Previously, we developed a tool called iPAGE to facilitate this analysis. However, both iPAGE and other similar tools implicitly assume that different genes have uniform gene-set membership. Our recent observations challenge this assumption, revealing that some genes form regulatory networks far more frequently than others. These technical biases and redundancies in gene-set annotations complicate the accurate inference of true regulatory relationships in specific contexts. To overcome these limitations, we introduced pyPAGE, an enhanced method that extends our information-theoretic framework by incorporating conditional mutual information to account for specific biases and artifacts. We applied pyPAGE to Alzheimer's disease, a neurodegenerative disorder marked by its complexity and characterized by progressive cognitive decline, memory loss, and behavioral changes. Pathological processes of Alzheimer's disease involve various cell types and diverse molecular mechanisms, which pose significant challenges for study. In our study we took advantage of our newly developed framework pyPAGE, performing analysis of gene expression changes at both tissue and cell type levels. This way we were able to describe regulation patterns of known regulators of Alzheimer's as well as several new ones. Additionally, we performed a cohort study of the association between the enrichment of the identified regulons and survival prognosis for patients, showing that increased activity of a subset of RBPs is positively associated with a lifespan. In total the results highlight the utility of pyPAGE and provide a valuable set of biomarkers for Alzheimer's disease. [ABSTRACT FROM AUTHOR]

Clustering is widely used in bioinformatics and many other fields, with applications from exploratory analysis to prediction. Many types of data have associated uncertainty or measurement error, but this is rarely used to inform the clustering. We present Dirichlet Process Mixtures with Uncertainty (DPMUnc), an extension of a Bayesian nonparametric clustering algorithm which makes use of the uncertainty associated with data points. We show that DPMUnc out-performs existing methods on simulated data. We cluster immune-mediated diseases (IMD) using GWAS summary statistics, which have uncertainty linked with the sample size of the study. DPMUnc separates autoimmune from autoinflammatory diseases and isolates other subgroups such as adult-onset arthritis. We additionally consider how DPMUnc can be used to cluster gene expression datasets that have been summarised using gene signatures. We first introduce a novel procedure for generating a summary of a gene signature on a dataset different to the one where it was discovered, which incorporates a measure of the variability in expression across signature genes within each individual. We summarise three public gene expression datasets containing patients with a range of IMD, using three relevant gene signatures. We find association between disease and the clusters returned by DPMUnc, with clustering structure replicated across the datasets. The significance of this work is two-fold. Firstly, we demonstrate that when data has associated uncertainty, this uncertainty should be used to inform clustering and we present a method which does this, DPMUnc. Secondly, we present a procedure for using gene signatures in datasets other than where they were originally defined. We show the value of this procedure by summarising gene expression data from patients with immune-mediated diseases using relevant gene signatures, and clustering these patients using DPMUnc. Author summary: Identifying groups of items that are similar to each other, a process called clustering, has a range of applications. For example, if patients split into two distinct groups this suggests that a disease may have subtypes which should be treated differently. Real data often has measurement error associated with it, but this error is frequently discarded by clustering methods. We propose a clustering method which makes use of the measurement error and use it to cluster diseases linked to the immune system. Gene expression datasets measure the activity level of all ∼20,000 genes in the human genome. We propose a procedure for summarising gene expression data using gene signatures, lists of genes produced by highly focused studies. For example, a study might list the genes which increase activity after exposure to a particular virus. The genes in the gene signature may not be as tightly correlated in a new dataset, and so our procedure measures the strength of the gene signature in the new dataset, effectively defining measurement error for the summary. We summarise gene expression datasets related to the immune system using relevant gene signatures and find that our method groups patients with the same disease. [ABSTRACT FROM AUTHOR]

MicroRNAs (miRNAs) are small non-coding RNAs that regulate gene expression post-transcriptionally. In animals, this regulation is achieved via base-pairing with partially complementary sequences on mainly 3' UTR region of messenger RNAs (mRNAs). Computational approaches that predict miRNA target interactions (MTIs) facilitate the process of narrowing down potential targets for experimental validation. The availability of new datasets of high-throughput, direct MTIs has led to the development of machine learning (ML) based methods for MTI prediction. To train an ML algorithm, it is beneficial to provide entries from all class labels (i.e., positive and negative). Currently, no high-throughput assays exist for capturing negative examples. Therefore, current ML approaches must rely on either artificially generated or inferred negative examples deduced from experimentally identified positive miRNA-target datasets. Moreover, the lack of uniform standards for generating such data leads to biased results and hampers comparisons between studies. In this comprehensive study, we collected methods for generating negative data for animal miRNAs–target interactions and investigated their impact on the classification of true human MTIs. Our study relies on training ML models on a fixed positive dataset in combination with different negative datasets and evaluating their intra- and cross-dataset performance. As a result, we were able to examine each method independently and evaluate ML models' sensitivity to the methodologies utilized in negative data generation. To achieve a deep understanding of the performance results, we analyzed unique features that distinguish between datasets. In addition, we examined whether one-class classification models that utilize solely positive interactions for training are suitable for the task of MTI classification. We demonstrate the importance of negative data in MTI classification, analyze specific methodological characteristics that differentiate negative datasets, and highlight the challenge of ML models generalizing interaction rules from training to testing sets derived from different approaches. This study provides valuable insights into the computational prediction of MTIs that can be further used to establish standards in the field. Author summary: Gene expression regulation is fundamental for all organisms' development, homeostasis, and environmental adaptation. microRNAs (miRNAs) play a central role in post-transcriptional gene regulation by binding to target mRNAs and repressing their translation or mediating their degradation. Technical challenges in experimental miRNA target identification led to growing interest in computational target prediction. While machine learning (ML) models have shown success in this area, they rely heavily on artificially generated negative examples due to limited experimental data. The diversity of methods for generating negative interactions and the lack of a uniform standardized approach introduce bias and hinder the comparison of results across different studies. In this study, we collected methods for generating negative data for animal miRNAs–target interactions and analyzed their impact on classifying true interactions in humans. Using an ML approach, we evaluated miRNA–target prediction performance within and across negative datasets. We also analyzed unique features distinguishing between the negative datasets to understand the performance results better. Our study shows that negative data is essential for accurately classifying miRNA–target interactions and that ML models struggle to apply what they have learned from the training set when faced with new data derived from different approaches. This study particularly appeals to researchers interested in miRNA–target classification. It emphasizes the need for standardized methods to enhance comparability between studies. [ABSTRACT FROM AUTHOR]

Finding communities in gene co-expression networks is a common first step toward extracting biological insight from these complex datasets. Most community detection algorithms expect genes to be organized into assortative modules, that is, groups of genes that are more associated with each other than with genes in other groups. While it is reasonable to expect that these modules exist, using methods that assume they exist a priori is risky, as it guarantees that alternative organizations of gene interactions will be ignored. Here, we ask: can we find meaningful communities without imposing a modular organization on gene co-expression networks, and how modular are these communities? For this, we use a recently developed community detection method, the weighted degree corrected stochastic block model (SBM), that does not assume that assortative modules exist. Instead, the SBM attempts to efficiently use all information contained in the co-expression network to separate the genes into hierarchically organized blocks of genes. Using RNA-seq gene expression data measured in two tissues derived from an outbred population of, we show that (a) the SBM is able to find ten times as many groups as competing methods, that (b) several of those gene groups are not modular, and that (c) the functional enrichment for non-modular groups is as strong as for modular communities. These results show that the transcriptome is structured in more complex ways than traditionally thought and that we should revisit the long-standing assumption that modularity is the main driver of the structuring of gene co-expression networks. Author summary: Understanding how genes work together is crucial for unraveling the biological processes underlying complex traits. To gain insight into these genetic interactions, researchers often analyze gene co-expression networks, in which genes are linked based on the similarity of their expression patterns among different individuals. Traditionally, it has been assumed that these networks are organized into distinct assortative modules, in which genes are more connected to each other within modules than between them. However, by using a novel non-parametric clustering approach called the Stochastic Block Model, we show that fruit fly transcriptomes contain not only assortative, modular gene clusters, but also functionally relevant non-modular clusters that would be overlooked by standard methods. This suggests that transcriptional networks may be more complex and diverse than previously thought. Our findings highlight the importance of using unbiased clustering techniques to fully capture the various architectures of gene co-expression networks and their potential biological significance. [ABSTRACT FROM AUTHOR]

Single-cell RNA sequencing (scRNA-seq) has become a popular experimental method to study variation of gene expression within a population of cells. However, obtaining an accurate picture of the diversity of distinct gene expression states that are present in a given dataset is highly challenging because the sparsity of the scRNA-seq data and its inhomogeneous measurement noise properties. Although a vast number of different methods is applied in the literature for clustering cells into subsets with 'similar' expression profiles, these methods generally lack rigorously specified objectives, involve multiple complex layers of normalization, filtering, feature selection, dimensionality-reduction, employ ad hoc measures of distance or similarity between cells, often ignore the known measurement noise properties of scRNA-seq measurements, and include a large number of tunable parameters. Consequently, it is virtually impossible to assign concrete biophysical meaning to the clusterings that result from these methods. Here we address the following problem: Given raw unique molecule identifier (UMI) counts of an scRNA-seq dataset, partition the cells into subsets such that the gene expression states of the cells in each subset are statistically indistinguishable, and each subset corresponds to a distinct gene expression state. That is, we aim to partition cells so as to maximally reduce the complexity of the dataset without removing any of its meaningful structure. We show that, given the known measurement noise structure of scRNA-seq data, this problem is mathematically well-defined and derive its unique solution from first principles. We have implemented this solution in a tool called Cellstates which operates directly on the raw data and automatically determines the optimal partition and cluster number, with zero tunable parameters. We show that, on synthetic datasets, Cellstates almost perfectly recovers optimal partitions. On real data, Cellstates robustly identifies subtle substructure within groups of cells that are traditionally annotated as a common cell type. Moreover, we show that the diversity of gene expression states that Cellstates identifies systematically depends on the tissue of origin and not on technical features of the experiments such as the total number of cells and total UMI count per cell. In addition to the Cellstates tool we also provide a small toolbox of software to place the identified cellstates into a hierarchical tree of higher-order clusters, to identify the most important differentially expressed genes at each branch of this hierarchy, and to visualize these results. Author summary: Single-cell RNA sequencing (scRNA-seq) has the promise to offer fundamental new insights into how gene expression is regulated but analyzing such data is very challenging because of its sparsity and heterogeneity in its measurement noise. For example, one common component of scRNA-seq analysis procedures is the clustering of groups of cells with similar gene expression, but current methods use complex ad hoc schemes that involve multiple layers of complex transformations of the data making it hard to interpret their results. Here we present a method that clusters cells so as maximally reduce the complexity of the data without removing any of its meaningful structure, by grouping only cells whose expression profiles are statistically indistinguishable. Importantly, we show that, given the known measurement noise structure of scRNA-seq data, this problem has a unique solution which we derive from first principles. We implemented this method in a tool called Cellstates which operates directly on the raw data with zero tunable parameters. We validate the power of Cellstates by showing it performs almost perfectly on synthetic data and outcompetes existing methods on real data consisting of known mixtures of cells. [ABSTRACT FROM AUTHOR]

Dimension reduction tools preserving similarity and graph structure such as t-SNE and UMAP can capture complex biological patterns in high-dimensional data. However, these tools typically are not designed to separate effects of interest from unwanted effects due to confounders. We introduce the partial embedding (PARE) framework, which enables removal of confounders from any distance-based dimension reduction method. We then develop partial t-SNE and partial UMAP and apply these methods to genomic and neuroimaging data. For lower-dimensional visualization, our results show that the PARE framework can remove batch effects in single-cell sequencing data as well as separate clinical and technical variability in neuroimaging measures. We demonstrate that the PARE framework extends dimension reduction methods to highlight biological patterns of interest while effectively removing confounding effects. Author summary: Powerful tools such as t-SNE and UMAP can be used to visualize biological patterns in medical data, including brain imaging and gene expression. However, these visuals can be influenced by unwanted patterns. We introduce the partial embedding (PARE) framework, which separates biological patterns from unwanted influences. We then develop partial t-SNE and partial UMAP and apply these methods to genomic and neuroimaging data. We first show that the PARE framework can remove technical artifacts in single-cell gene expression data. Then, we show that PAREs can reduce unwanted scanner-related bias in brain imaging data. We demonstrate that the PARE framework highlights biological patterns of interest while effectively removing unwanted effects. [ABSTRACT FROM AUTHOR]

Transcription factors (TFs) regulate the process of transcription through the modulation of different kinetic steps. Although models can often describe the observed transcriptional output of a measured gene, predicting a TFs role on a given promoter requires an understanding of how the TF alters each step of the transcription process. In this work, we use a simple model of transcription to assess the role of promoter identity, and the degree to which TFs alter binding of RNAP (stabilization) and initiation of transcription (acceleration) on three primary characteristics: the range of steady-state regulation, cell-to-cell variability in expression, and the dynamic response time of a regulated gene. We find that steady state regulation and the response time of a gene behave uniquely for TFs that regulate incoherently, i.e that speed up one step but slow the other. We also find that incoherent TFs have dynamic implications, with one type of incoherent mode configuring the promoter to respond more slowly at intermediate TF concentrations. We also demonstrate that the noise of gene expression for these TFs is sensitive to promoter strength, with a distinct non-monotonic profile that is apparent under stronger promoters. Taken together, our work uncovers the coupling between promoters and TF regulatory modes with implications for understanding natural promoters and engineering synthetic gene circuits with desired expression properties. Author summary: In this study, we explored the regulatory roles of TFs in gene expression by developing a model that considers the effects of two parameters: stabilization (interactions that modify binding of RNA polymerase to the promoter) and acceleration (interactions that modify the rate of transcription initiation). By applying this model, we evaluated how these factors influence gene expression dynamics, noise, and response times. We show that different combinations of stabilization and acceleration can lead to a rich diversity of regulatory properties. The findings offer valuable insights into the complex interactions between TFs and promoters, underscores the need to characterize TFs in terms of their regulatory modes, and potential guides the engineering of synthetic circuits with desired expression characteristics. [ABSTRACT FROM AUTHOR]

In embryonic development and organogenesis, cells sharing identical genetic codes acquire diverse gene expression states in a highly reproducible spatial distribution, crucial for multicellular formation and quantifiable through positional information. To understand the spontaneous growth of complexity, we constructed a one-dimensional division-decision model, simulating the growth of cells with identical genetic networks from a single cell. Our findings highlight the pivotal role of cell division in providing positional cues, escorting the system toward states rich in information. Moreover, we pinpointed lateral inhibition as a critical mechanism translating spatial contacts into gene expression. Our model demonstrates that the spatial arrangement resulting from cell division, combined with cell lineages, imparts positional information, specifying multiple cell states with increased complexity—illustrated through examples in C.elegans. This study constitutes a foundational step in comprehending developmental intricacies, paving the way for future quantitative formulations to construct synthetic multicellular patterns. Author summary: Embryonic development shapes our bodies from a single cell, determining the placement of the head and tail. But how do cells, all sharing the same genetic code, precisely know what to become? Our mathematical model cracks this code. Envision your information being provided by your neighbors and your 'mom'—the division lineage. Then, appropriate regulatory networks (such as lateral inhibition, the most effective network motif) transform this information into diverse yet robust gene expressions. This math model helps us see the rules behind spontaneously growing complexity, guiding us to create new patterns of cells. It's a big step in understanding how bodies form and opens doors to building cool new structures. [ABSTRACT FROM AUTHOR]

The activation levels of biologically significant gene sets are emerging tumor molecular markers and play an irreplaceable role in the tumor research field; however, web-based tools for prognostic analyses using it as a tumor molecular marker remain scarce. We developed a web-based tool PESSA for survival analysis using gene set activation levels. All data analyses were implemented via R. Activation levels of The Molecular Signatures Database (MSigDB) gene sets were assessed using the single sample gene set enrichment analysis (ssGSEA) method based on data from the Gene Expression Omnibus (GEO), The Cancer Genome Atlas (TCGA), The European Genome-phenome Archive (EGA) and supplementary tables of articles. PESSA was used to perform median and optimal cut-off dichotomous grouping of ssGSEA scores for each dataset, relying on the survival and survminer packages for survival analysis and visualisation. PESSA is an open-access web tool for visualizing the results of tumor prognostic analyses using gene set activation levels. A total of 238 datasets from the GEO, TCGA, EGA, and supplementary tables of articles; covering 51 cancer types and 13 survival outcome types; and 13,434 tumor-related gene sets are obtained from MSigDB for pre-grouping. Users can obtain the results, including Kaplan–Meier analyses based on the median and optimal cut-off values and accompanying visualization plots and the Cox regression analyses of dichotomous and continuous variables, by selecting the gene set markers of interest. PESSA (https://smuonco.shinyapps.io/PESSA/ OR http://robinl-lab.com/PESSA) is a large-scale web-based tumor survival analysis tool covering a large amount of data that creatively uses predefined gene set activation levels as molecular markers of tumors. Author summary: The activation levels of biologically significant gene sets are emerging tumor molecular markers and play an irreplaceable role in the tumor research field; however, web-based tools for prognostic analyses using it as a tumor molecular marker remain scarce. PESSA is an open-access web tool for visualizing the results of tumor prognostic analyses using gene set activation levels. A total of 238 datasets from the GEO, TCGA, EGA, and supplementary tables of articles; covering 51 cancer types and 13 survival outcome types; and 13,434 tumor-related gene sets are obtained from MSigDB for pre-grouping. Users can obtain the results, including Kaplan–Meier analyses based on the median and optimal cut-off values and visualization plots and the Cox regression analyses of dichotomous and continuous variables, by selecting the gene set markers of interest. Freely available on the web at https://smuonco.shinyapps.io/PESSA/ OR http://robinl-lab.com/PESSA. Website implemented in R Shiny, with major browsers supported. [ABSTRACT FROM AUTHOR]

Cancer development is driven by an accumulation of a small number of driver genetic mutations that confer the selective growth advantage to the cell, while most passenger mutations do not contribute to tumor progression. The identification of these driver genes responsible for tumorigenesis is a crucial step in designing effective cancer treatments. Although many computational methods have been developed with this purpose, the majority of existing methods solely provided a single driver gene list for the entire cohort of patients, ignoring the high heterogeneity of driver events across patients. It remains challenging to identify the personalized driver genes. Here, we propose a novel method (PDRWH), which aims to prioritize the mutated genes of a single patient based on their impact on the abnormal expression of downstream genes across a group of patients who share the co-mutation genes and similar gene expression profiles. The wide experimental results on 16 cancer datasets from TCGA showed that PDRWH excels in identifying known general driver genes and tumor-specific drivers. In the comparative testing across five cancer types, PDRWH outperformed existing individual-level methods as well as cohort-level methods. Our results also demonstrated that PDRWH could identify both common and rare drivers. The personalized driver profiles could improve tumor stratification, providing new insights into understanding tumor heterogeneity and taking a further step toward personalized treatment. We also validated one of our predicted novel personalized driver genes on tumor cell proliferation by vitro cell-based assays, the promoting effect of the high expression of Low-density lipoprotein receptor-related protein 1 (LRP1) on tumor cell proliferation. Author summary: In this study, using the TCGA dataset studies as benchmark datasets, we explored the application of the commonality among patients of the same cancer type in personalized driver gene prediction. We proposed a hypergraph model and a generalized random walk method to rank the mutated genes of a patient based on their impact on the abnormal expression of downstream genes in a group of samples rather than an individual sample. Following the extensive experimental results on 16 cancer datasets and the comparative analysis across five cancer types, we have observed that the PDRWH method exhibits remarkable effectiveness in identifying known general driver genes and tumor-specific driver genes. In a few words, our method can provide a more accurate personalized catalog of driver mutations for each patient, and the predicted personalized driver genes can be applied to improve tumor stratification. It can also provide oncologists with a reliable candidate gene list to assist treatment decisions, thus potentially promoting the development of personalized medicine. [ABSTRACT FROM AUTHOR]

Existing studies have shown that the abnormal expression of microRNAs (miRNAs) usually leads to the occurrence and development of human diseases. Identifying disease-related miRNAs contributes to studying the pathogenesis of diseases at the molecular level. As traditional biological experiments are time-consuming and expensive, computational methods have been used as an effective complement to infer the potential associations between miRNAs and diseases. However, most of the existing computational methods still face three main challenges: (i) learning of high-order relations; (ii) insufficient representation learning ability; (iii) importance learning and integration of multi-view embedding representation. To this end, we developed a HyperGraph Contrastive Learning with view-aware Attention Mechanism and Integrated multi-view Representation (HGCLAMIR) model to discover potential miRNA-disease associations. First, hypergraph convolutional network (HGCN) was utilized to capture high-order complex relations from hypergraphs related to miRNAs and diseases. Then, we combined HGCN with contrastive learning to improve and enhance the embedded representation learning ability of HGCN. Moreover, we introduced view-aware attention mechanism to adaptively weight the embedded representations of different views, thereby obtaining the importance of multi-view latent representations. Next, we innovatively proposed integrated representation learning to integrate the embedded representation information of multiple views for obtaining more reasonable embedding information. Finally, the integrated representation information was fed into a neural network-based matrix completion method to perform miRNA-disease association prediction. Experimental results on the cross-validation set and independent test set indicated that HGCLAMIR can achieve better prediction performance than other baseline models. Furthermore, the results of case studies and enrichment analysis further demonstrated the accuracy of HGCLAMIR and unconfirmed potential associations had biological significance. Author summary: Considerable studies have demonstrated that the dysregulation of miRNAs is closely related to human diseases. Therefore, inferring unconfirmed associations between miRNAs and diseases is helpful for disease diagnosis and treatment. Numerous computational models have been proposed to discover potential miRNA-disease associations on a large scale, which can accelerate the understanding of disease pathogenesis. We constructed a HGCLAMIR model to identify miRNA-disease associations through hypergraph convolutional network with contrastive learning, view-aware attention mechanism and integrated representation learning. The 5-fold cross-validation and independent testing were performed to evaluate the performance of HGCLAMIR, which was better than ten baseline models. In addition, we carried out case studies on breast neoplasms and lung neoplasms, showing that 49 and 48 of the top 50 candidate miRNAs were confirmed by experimental reports. In summary, HGCLAMIR could be considered as an effective and accurate model for predicting the associations between miRNAs and diseases. [ABSTRACT FROM AUTHOR]

Prioritization or ranking of different cell types in a Single-cell RNA Sequencing (scRNA-Seq) framework can be performed in a variety of ways, some of these include: i) obtaining an indication of the proportion of cell types between the different conditions under study, ii) counting the number of differentially expressed genes (DEGs) between cell types and conditions in the experiment or, iii) prioritizing cell types based on prior knowledge about the conditions under study (i.e., a specific disease). These methods have drawbacks and limitations thus novel methods for improving cell ranking are required. Here we present a novel methodology that exploits prior knowledge in combination with expert-user information to accentuate cell types from a scRNA-seq analysis that yield the most biologically meaningful results with respect to a disease under study. Our methodology allows for ranking and prioritization of cell-types based on how well their expression profiles relate to the molecular mechanisms and drugs associated with a disease. Molecular mechanisms, as well as drugs, are incorporated as prior knowledge in a standardized, structured manner. Cell-types are then ranked/prioritized based on how well results from data-driven analysis of scRNA-seq data match the predefined prior knowledge. In additional cell-cell communication perturbations between disease and control networks are used to further prioritize/rank cell-types. Our methodology has substantial advantages to more traditional cell ranking techniques and provides an informative complementary methodology that utilizes prior knowledge in a rapid and automated manner, that has previously not been attempted by other studies. The current methodology is also implemented as an R package entitled Single Cell Ranking Analysis Toolkit (scRANK) and is available for download and installation via GitHub (https://#hub.com/aoulas/scRANK) Author summary: Single-cell RNA Sequencing (scRNA-Seq) provides an additive resolution down to the cellular level that was previously not available from traditional "Bulk" RNA sequencing experiments. However, it is often difficult to prioritize the specific cell-types which are primarily responsible for the cause of a disease. This work presents a novel methodology that utilizes predefined prior knowledge information to highlight informative cell types from a scRNA-seq analysis. Our methodology allows for ranking and prioritization of cell-types based on how well predefined prior knowledge, in the form of pathways and drugs, are mapped to the results obtained from a basic analysis of a scRNA-Seq dataset. Prior knowledge is incorporated in a standardized, structured manner, whereby a checklist is attained by querying the human disease database called MalaCards. This checklist comes in the form of pathways and drugs associated with the disease. The expert-user is prompted to "edit" the prior knowledge checklist by removing or adding terms from the list of predefined terms. Our methodology also utilizes cell-cell communication perturbations between disease and control networks to further prioritize/rank cell-types. We believe, this methodology may offer substantial advantages to existing cell prioritizing or ranking techniques and provide a rapid and automated approach that further compliments other methods of cell prioritization in a scRNA-Seq framework. [ABSTRACT FROM AUTHOR]

Recent advances in single-cell sequencing technology have provided opportunities for mathematical modeling of dynamic developmental processes at the single-cell level, such as inferring developmental trajectories. Optimal transport has emerged as a promising theoretical framework for this task by computing pairings between cells from different time points. However, optimal transport methods have limitations in capturing nonlinear trajectories, as they are static and can only infer linear paths between endpoints. In contrast, stochastic differential equations (SDEs) offer a dynamic and flexible approach that can model non-linear trajectories, including the shape of the path. Nevertheless, existing SDE methods often rely on numerical approximations that can lead to inaccurate inferences, deviating from true trajectories. To address this challenge, we propose a novel approach combining forward-backward stochastic differential equations (FBSDE) with a refined approximation procedure. Our FBSDE model integrates the forward and backward movements of two SDEs in time, aiming to capture the underlying dynamics of single-cell developmental trajectories. Through comprehensive benchmarking on multiple scRNA-seq datasets, we demonstrate the superior performance of FBSDE compared to other methods, highlighting its efficacy in accurately inferring developmental trajectories. Author summary: In this study, we address the challenge of modeling the trajectories of single-cell gene expression over time. A "trajectory" in this context refers to the dynamic evolution of gene expression within a single cell, and this challenge is exacerbated by the fact that the gene expression of sequenced cells cannot be observed in subsequent time points. To overcome this limitation, we employ innovative stochastic differential equations (SDEs), a technique recently utilized in trajectory modeling studies, to more accurately capture the dynamic evolution of single-cell gene expression. Rigorously tested on multiple single-cell RNA sequencing datasets, our enhanced SDE system demonstrates promising performance in capturing these intricate gene expression dynamics, thereby advancing our capacity to glean critical insights across various biological contexts. [ABSTRACT FROM AUTHOR]

Determining the general laws between evolution and development is a fundamental biological challenge. Developmental hourglasses have attracted increased attention as candidates for such laws, but the necessity of their emergence remains elusive. We conducted evolutionary simulations of developmental processes to confirm the emergence of the developmental hourglass and unveiled its establishment. We considered organisms consisting of cells containing identical gene networks that control morphogenesis and evolved them under selection pressure to induce more cell types. By computing the similarity between the spatial patterns of gene expression of two species that evolved from a common ancestor, a developmental hourglass was observed, that is, there was a correlation peak in the intermediate stage of development. The fraction of pleiotropic genes increased, whereas the variance in individuals decreased, consistent with previous experimental reports. Reduction of the unavoidable variance by initial or developmental noise, essential for survival, was achieved up to the hourglass bottleneck stage, followed by diversification in developmental processes, whose timing is controlled by the slow expression dynamics conserved among organisms sharing the hourglass. This study suggests why developmental hourglasses are observed within a certain phylogenetic range of species. Author summary: Understanding the intriguing relationship between development and evolution in multicellular organisms has long been a challenge in biology. A recent hypothesis called the developmental hourglass proposes that there is a conserved middle stage during development across species of the same animal group. Despite growing evidence supporting this hypothesis, the underlying mechanisms and reasons for its emergence have remained elusive due to limited experimental data. To address this gap, we employed numerical evolution of gene regulation networks controlling pattern formation. Remarkably, our simulations revealed that species that diverged relatively recently in phylogeny displayed the highest similarity during the middle stage of development, which gradually diminished as they diverged further phylogenetically. Our findings satisfied not only the criteria of the developmental hourglass but also confirmed several essential characteristics of the developmental hourglass reported in recent experiments. Through theoretical analysis, we further demonstrated that the emergence of the developmental hourglass could be attributed to the acquisition of genes that change slowly and govern developmental processes, which also foster the robustness of development. By integrating computational simulations, theoretical insights, and previous experimental evidence, our study thus provides a comprehensive understanding of the developmental hourglass, which will unravel the intricate relationship between development and evolution. [ABSTRACT FROM AUTHOR]

Spatial transcriptomic (ST) clustering employs spatial and transcription information to group spots spatially coherent and transcriptionally similar together into the same spatial domain. Graph convolution network (GCN) and graph attention network (GAT), fed with spatial coordinates derived adjacency and transcription profile derived feature matrix are often used to solve the problem. Our proposed method STGIC (spatial transcriptomic clustering with graph and image convolution) is designed for techniques with regular lattices on chips. It utilizes an adaptive graph convolution (AGC) to get high quality pseudo-labels and then resorts to dilated convolution framework (DCF) for virtual image converted from gene expression information and spatial coordinates of spots. The dilation rates and kernel sizes are set appropriately and updating of weight values in the kernels is made to be subject to the spatial distance from the position of corresponding elements to kernel centers so that feature extraction of each spot is better guided by spatial distance to neighbor spots. Self-supervision realized by Kullback–Leibler (KL) divergence, spatial continuity loss and cross entropy calculated among spots with high confidence pseudo-labels make up the training objective of DCF. STGIC attains state-of-the-art (SOTA) clustering performance on the benchmark dataset of 10x Visium human dorsolateral prefrontal cortex (DLPFC). Besides, it's capable of depicting fine structures of other tissues from other species as well as guiding the identification of marker genes. Also, STGIC is expandable to Stereo-seq data with high spatial resolution. Author summary: Spatial transcriptomics detect gene transcription profile with high spatial resolution even to sub-cellular level, which is very helpful to characterize organization and architecture of tissues and find marker genes corresponding to sub-structures. Therefore, it is much more informative than traditional transcriptomics for interpreting the biological processes underlying tissue development and disease progression. Clustering to clarify each spatial domain of tissues in which cells present similar transcription profile and histology is a primary task for analyzing spatial transcriptomics data. To solve the important problem, we propose the method STGIC combining deep learning skills in both graph and image, since we notice that image methods tend to display good performances on some samples in benchmark datasets which existing graph methods are not good at, and vice versa. Besides, we substitute the vanilla form of graph convolution frequently used in this field with the one which can find the optimal order of neighbor involved in feature aggregation adaptively for different graphs. STGIC displays high clustering performance and identifies spatial domains which can be used to depict fine-grained structure and help ascertain marker genes of different tissues from different species sequenced with 10x Visum and Stereo-seq techniques. [ABSTRACT FROM AUTHOR]

Single-cell RNA sequencing (scRNA-seq) technology produces an unprecedented resolution at the level of a unique cell, raising great hopes in medicine. Nevertheless, scRNA-seq data suffer from high variations due to the experimental conditions, called batch effects, preventing any aggregated downstream analysis. Adversarial Information Factorization provides a robust batch-effect correction method that does not rely on prior knowledge of the cell types nor a specific normalization strategy while being adapted to any downstream analysis task. It compares to and even outperforms state-of-the-art methods in several scenarios: low signal-to-noise ratio, batch-specific cell types with few cells, and a multi-batches dataset with imbalanced batches and batch-specific cell types. Moreover, it best preserves the relative gene expression between cell types, yielding superior differential expression analysis results. Finally, in a more complex setting of a Leukemia cohort, our method preserved most of the underlying biological information for each patient while aligning the batches, improving the clustering metrics in the aggregated dataset. Author summary: Single-cell RNA sequencing captures the signal of individual cells, allowing a finer resolution than bulk sequencing, which is particularly important for studies comprising rare populations like tumor heterogeneity or lineage tracing studies. However, it is sensitive to the experimental conditions, which induce a bias in the data, called batch effects. Those technical variations hinder any aggregated analysis, limiting scRNA-seq to individual trials. To address this issue, we developed a novel Deep-Learning method called Adversarial Information Factorization, which aims at factorizing the batch effects from the biological signal to align the individual trials for downstream aggregated analysis. The model is trained to learn the batch-conditional cells' distributions and then corrects batch effects by projecting all cells onto the same batch distribution. [ABSTRACT FROM AUTHOR]

In comparisons between mutant and wild-type genotypes, transcriptome analysis can reveal the direct impacts of a mutation, together with the homeostatic responses of the biological system. Recent studies have highlighted that, when the effects of homozygosity for recessive mutations are studied in non-isogenic backgrounds, genes located proximal to the mutation on the same chromosome often appear over-represented among those genes identified as differentially expressed (DE). One hypothesis suggests that DE genes chromosomally linked to a mutation may not reflect functional responses to the mutation but, instead, result from an unequal distribution of expression quantitative trait loci (eQTLs) between sample groups of mutant or wild-type genotypes. This is problematic because eQTL expression differences are difficult to distinguish from genes that are DE due to functional responses to a mutation. Here we show that chromosomally co-located differentially expressed genes (CC-DEGs) are also observed in analyses of dominant mutations in heterozygotes. We define a method and a metric to quantify, in RNA-sequencing data, localised differential allelic representation (DAR) between those sample groups subjected to differential expression analysis. We show how the DAR metric can predict regions prone to eQTL-driven differential expression, and how it can improve functional enrichment analyses through gene exclusion or weighting-based approaches. Advantageously, this improved ability to identify probable eQTLs also reveals examples of CC-DEGs that are likely to be functionally related to a mutant phenotype. This supports a long-standing prediction that selection for advantageous linkage disequilibrium influences chromosome evolution. By comparing the genomes of zebrafish (Danio rerio) and medaka (Oryzias latipes), a teleost with a conserved ancestral karyotype, we find possible examples of chromosomal aggregation of CC-DEGs during evolution of the zebrafish lineage. Our method for DAR analysis requires only RNA-sequencing data, facilitating its application across new and existing datasets. Author summary: Many human-relevant diseases result from genetic mutations that disrupt cellular functions. We can model these mutations in other organisms (e.g. mouse, zebrafish) and employ gene expression analysis (transcriptomics) to determine how mutations directly affect cells and how cells adjust expression of their genes to compensate for these mutations. In our transcriptome analyses of dominant disease-causative mutations in zebrafish, we identified an interesting phenomenon where a disproportionate number of differentially expressed genes reside on the same chromosome as a mutated gene. Here, we provide strong evidence supporting that the differential expression of some of these chromosomally co-located genes is not due to the mutation but is due to differential segregation of gene alleles with innately different expression levels (i.e. expression quantitative trait loci, eQTLs). We have developed a procedure to measure the likelihood of differential gene expression being due to an eQTL. This allows us to compensate for the presence of such eQTLs in bioinformatic analyses. Our procedure, Differential Allelic Representation (DAR) analysis, revealed evidence for aggregation of genes with related functions on the same chromosome over evolutionary timescales. DAR analysis allows disentanglement of eQTLs from mutation-dependent gene expression responses, thereby permitting more comprehensive investigation of transcriptome data. [ABSTRACT FROM AUTHOR]

Generalist microbes have adapted to a multitude of environmental stresses through their integrated stress response system. Individual stress responses have been quantified by E. coli metabolism and expression (ME) models under thermal, oxidative and acid stress, respectively. However, the systematic quantification of cross-stress & cross-talk among these stress responses remains lacking. Here, we present StressME: the unified stress response model of E. coli combining thermal (FoldME), oxidative (OxidizeME) and acid (AcidifyME) stress responses. StressME is the most up to date ME model for E. coli and it reproduces all published single-stress ME models. Additionally, it includes refined rate constants to improve prediction accuracy for wild-type and stress-evolved strains. StressME revealed certain optimal proteome allocation strategies associated with cross-stress and cross-talk responses. These stress-optimal proteomes were shaped by trade-offs between protective vs. metabolic enzymes; cytoplasmic vs. periplasmic chaperones; and expression of stress-specific proteins. As StressME is tuned to compute metabolic and gene expression responses under mild acid, oxidative, and thermal stresses, it is useful for engineering and health applications. The modular design of our open-source package also facilitates model expansion (e.g., to new stress mechanisms) by the computational biology community. Author summary: A fundamental understanding of multi-stress adaptation in E.coli has potential industrial relevance. While individual stress responses have been quantified through the protein regulatory network in E.coli, the systematic quantification of the cross-stress & cross-talk among stress responses remains lacking. Here, we develop a new modeling pipeline by which thermal, oxidative and acid stress response can be coupled to each other, and the metabolic activities, protein and metabolic flux redistribution due to cross-stress & cross-talk can be quantified. We optimize the effective rate constants in the integrated model. We then confirm the model robustness by validating against the published data under single stress. Finally, we use the model to characterize the cross-adaptation between protective and catalytic proteins as well as between chaperones present in different cellular compartments. We find effective cross-protection against cross stress by adapting the E.coli cells to the thermal stress first. We also indicate the presence of cross-talk through trade-offs by which the cell may refuse to give up more protein allocation away from one stress response to the other, because doing so would decrease stress tolerance further. The single stress plug-in design makes the model build-up pipeline flexible and expandable, allowing incorporation of more stressors into the model architecture for industrial applications. [ABSTRACT FROM AUTHOR]

Super enhancers (SE), large genomic elements that activate transcription and drive cell identity, have been found with cancer-specific gene regulation in human cancers. Recent studies reported the importance of understanding the cooperation and function of SE internal components, i.e., the constituent enhancers (CE). However, there are no pan-cancer studies to identify cancer-specific SE signatures at the constituent level. Here, by revisiting pan-cancer SE activities with H3K27Ac ChIP-seq datasets, we report fingerprint SE signatures for 28 cancer types in the NCI-60 cell panel. We implement a mixture model to discriminate active CEs from inactive CEs by taking into consideration ChIP-seq variabilities between cancer samples and across CEs. We demonstrate that the model-based estimation of CE states provides improved functional interpretation of SE-associated regulation. We identify cancer-specific CEs by balancing their active prevalence with their capability of encoding cancer type identities. We further demonstrate that cancer-specific CEs have the strongest per-base enhancer activities in independent enhancer sequencing assays, suggesting their importance in understanding critical SE signatures. We summarize fingerprint SEs based on the cancer-specific statuses of their component CEs and build an easy-to-use R package to facilitate the query, exploration, and visualization of fingerprint SEs across cancers. Author summary: Super enhancers are large genomic elements comprised of multiple enhancers working together to drive gene transcription. They play a crucial role in defining cell identity and act as drivers of oncogenic gene expression in cancer cells. Characterizing cancer-specific super enhancer signatures can reveal transcriptional deregulation associated with cell origin and malignant transformation. Here, we generated a high-resolution fingerprint of super enhancers across 60 cancer cell lines through statistical modeling of both active and inactive components inside super enhancers. Our study revealed that cancer-specific super enhancer components are highly informative in delineating the identity of cancer cells. Our findings further revealed that cancer-specific active components exhibit stronger enhancer activities compared to non-cancer-specific components, suggesting the importance of studying the functional divergence inside super enhancers across different cancer types. Finally, we generated a database of cancer-specific super enhancer signatures for 28 cancer types with a companion computational tool to facilitate the query, exploration, and visualization of these signatures across cancers. [ABSTRACT FROM AUTHOR]

The transcriptional regulatory network (TRN) of E. coli consists of thousands of interactions between regulators and DNA sequences. Regulons are typically determined either from resource-intensive experimental measurement of functional binding sites, or inferred from analysis of high-throughput gene expression datasets. Recently, independent component analysis (ICA) of RNA-seq compendia has shown to be a powerful method for inferring bacterial regulons. However, it remains unclear to what extent regulons predicted by ICA structure have a biochemical basis in promoter sequences. Here, we address this question by developing machine learning models that predict inferred regulon structures in E. coli based on promoter sequence features. Models were constructed successfully (cross-validation AUROC > = 0.8) for 85% (40/47) of ICA-inferred E. coli regulons. We found that: 1) The presence of a high scoring regulator motif in the promoter region was sufficient to specify regulatory activity in 40% (19/47) of the regulons, 2) Additional features, such as DNA shape and extended motifs that can account for regulator multimeric binding, helped to specify regulon structure for the remaining 60% of regulons (28/47); 3) investigating regulons where initial machine learning models failed revealed new regulator-specific sequence features that improved model accuracy. Finally, we found that strong regulatory binding sequences underlie both the genes shared between ICA-inferred and experimental regulons as well as genes in the E. coli core pan-regulon of Fur. This work demonstrates that the structure of ICA-inferred regulons largely can be understood through the strength of regulator binding sites in promoter regions, reinforcing the utility of top-down inference for regulon discovery. Author summary: The Transcriptional Regulatory Network (TRN) of bacteria is a gene expression control system largely governed by transcription factors binding to promoter regions of the genome, forming a connected network of regulons. While experimental determination of functional transcription factor binding sites is laborious, inference of regulons from large-scale gene expression datasets has proven to be a powerful alternative. Here, we seek to determine whether inferred regulons have a biochemical basis in terms of binding sites sequences present at gene promoters. We use machine learning to build quantitative models relating binding site features to gene regulation. We find that the majority of data-inferred regulons can be tied to binding site sequences at gene promoters, supporting the biological reality of inferred regulons. Additionally, these inferred regulons sometimes carried additional sequence context that is not accounted for by the transcription factor binding motif extracted from experimental binding sites. An understanding of how regulons are encoded in the genome will empower sequence analysis and synthetic biology applications. [ABSTRACT FROM AUTHOR]

Motivation: Transcriptional regulation is performed by transcription factors (TF) binding to DNA in context-dependent regulatory regions and determines the activation or inhibition of gene expression. Current methods of transcriptional regulatory circuits inference, based on one or all of TF, regions and genes activity measurements require a large number of samples for ranking the candidate TF-gene regulation relations and rarely predict whether they are activations or inhibitions. We hypothesize that transcriptional regulatory circuits can be inferred from fewer samples by (1) fully integrating information on TF binding, gene expression and regulatory regions accessibility, (2) reducing data complexity and (3) using biology-based likelihood constraints to determine the global consistency between a candidate TF-gene relation and patterns of genes expressions and region activations, as well as qualify regulations as activations or inhibitions. Results: We introduce Regulus, a method which computes TF-gene relations from gene expressions, regulatory region activities and TF binding sites data, together with the genomic locations of all entities. After aggregating gene expressions and region activities into patterns, data are integrated into a RDF (Resource Description Framework) endpoint. A dedicated SPARQL (SPARQL Protocol and RDF Query Language) query retrieves all potential relations between expressed TF and genes involving active regulatory regions. These TF-region-gene relations are then filtered using biological likelihood constraints allowing to qualify them as activation or inhibition. Regulus provides signed relations consistent with public databases and, when applied to biological data, identifies both known and potential new regulators. Regulus is devoted to context-specific transcriptional circuits inference in human settings where samples are scarce and cell populations are closely related, using discretization into patterns and likelihood reasoning to decipher the most robust regulatory relations. Author summary: Gene expression regulation is based on the activity of specialized regulatory proteins called transcription factors (TFs) which can bind DNA at specific sequences. Understanding the regulatory relations between TFs and genes in humans is fundamental in personalized clinical settings, to better decipher the pathological mechanisms and to identify new therapeutic solutions. However, finding the main regulators of such systems is usually difficult, due to the scarcity of available samples and the biological closeness of the studied cell types. To overcome these issues, we introduce a new tool called Regulus. We use information from genes and TFs expression, regulatory regions activity and TF binding sites occurrences to compute TF-gene relations. We then apply a likelihood reasoning step, based on the biological knowledge of transcriptional regulation mechanisms, to select the most probable relations and assign them a function as activation or inhibition. Finally, we reduce the potential TFs list by a specificity / coverage filter and we annotate it according to existing literature. By testing Regulus onlarge-scale biological datasets, each describing four biological contexts, we show that this tool is able to i) identify both known and undescribed regulators consistent with all the gene expression and region accessibility constraints in each biological context, ii) include low expressed genes in its relations and iii) considerably limit the space of putative TF-gene relations. [ABSTRACT FROM AUTHOR]

The effects of transcription factor binding sites (TFBSs) on the activity of a cis-regulatory element (CRE) depend on the local sequence context. In rod photoreceptors, binding sites for the transcription factor (TF) Cone-rod homeobox (CRX) occur in both enhancers and silencers, but the sequence context that determines whether CRX binding sites contribute to activation or repression of transcription is not understood. To investigate the context-dependent activity of CRX sites, we fit neural network-based models to the activities of synthetic CREs composed of photoreceptor TFBSs. The models revealed that CRX binding sites consistently make positive, independent contributions to CRE activity, while negative homotypic interactions between sites cause CREs composed of multiple CRX sites to function as silencers. The effects of negative homotypic interactions can be overcome by the presence of other TFBSs that either interact cooperatively with CRX sites or make independent positive contributions to activity. The context-dependent activity of CRX sites is thus determined by the balance between positive heterotypic interactions, independent contributions of TFBSs, and negative homotypic interactions. Our findings explain observed patterns of activity among genomic CRX-bound enhancers and silencers, and suggest that enhancers may require diverse TFBSs to overcome negative homotypic interactions between TFBSs. Author summary: Transcription factors control gene expression in different cell types by binding to sites in regulatory DNA. The same transcription factor when bound at different DNA sites will have different effects on gene expression, but how a single factor can produce divergent effects is unclear. The photoreceptor transcription factor CRX activates expression from regulatory DNA that harbors few copies of a CRX binding site, while it represses expression when many binding site copies are present. We modeled how the number and arrangement of binding sites for CRX and other factors affect gene expression, using data from libraries of synthetic regulatory DNA elements. The model shows that individual transcription factor binding sites increase expression on their own, but interactions between multiple copies of the same site decrease expression. Our results generalize across transcription factors and tissues, suggesting that this is a general principle that might help explain differing patterns of expression across tissues. The model explains how interactions between binding sites allow a single transcription factor to have contrasting effects on gene expression in the same cell type. [ABSTRACT FROM AUTHOR]