Differential correlation for sequencing data

The transcriptional regulatory network modulating human trophoblast stem cells to extravillous trophoblast differentiation

During human pregnancy, extravillous trophoblasts play crucial roles in placental invasion into the maternal decidua and spiral artery remodeling. However, regulatory factors and their action mechanisms modulating human extravillous trophoblast specification have been unknown. By analyzing dynamic changes in transcriptome and enhancer profile during human trophoblast stem cell to extravillous trophoblast differentiation, we define stage-specific regulators, including an early-stage transcription factor, TFAP2C, and multiple late-stage transcription factors. Loss-of-function studies confirm the requirement of all transcription factors identified for adequate differentiation, and we reveal that the dynamic changes in the levels of TFAP2C are essential. Notably, TFAP2C pre-occupies the regulatory elements of the inactive extravillous trophoblast-active genes during the early stage of differentiation, and the late-stage transcription factors directly activate extravillous trophoblast-active genes, including themselves as differentiation further progresses, suggesting sequential actions of transcription factors assuring differentiation. Our results reveal stage-specific transcription factors and their inter-connected regulatory mechanisms modulating extravillous trophoblast differentiation, providing a framework for understanding early human placentation and placenta-related complications.

Correlations for untargeted GC × GC-HRTOF-MS metabolomics of colorectal cancer

INTRODUCTION

Modern comprehensive instrumentations provide an unprecedented coverage of complex matrices in the form of high-dimensional, information rich data sets.

OBJECTIVES

In addition to the usual biomarker research that focuses on the detection of the studied condition, we aimed to define a proper strategy to conduct a correlation analysis on an untargeted colorectal cancer case study with a data set of 102 variables corresponding to metabolites obtained from serum samples analyzed with comprehensive two-dimensional gas chromatography coupled to high-resolution time-of-flight mass spectrometry (GC × GC-HRTOF-MS). Indeed, the strength of association existing between the metabolites contains potentially valuable information about the molecular mechanisms involved and the underlying metabolic network associated to a global perturbation, at no additional analytical effort.

METHODS

Following Anscombe's quartet, we took particular attention to four main aspects. First, the presence of non-linear relationships through the comparison of parametric and non-parametric correlation coefficients: Pearson's r, Spearman's rho, Kendall's tau and Goodman-Kruskal's gamma. Second, the visual control of the detected associations through scatterplots and their associated regressions and angles. Third, the effect and handling of atypical samples and values. Fourth, the role of the precision of the data on the attribution of the ranks through the presence of ties.

RESULTS

Kendall's tau was found the method of choice for the data set at hand. Its application highlighted 17 correlations significantly altered in the active state of colorectal cancer (CRC) in comparison to matched healthy controls (HC), from which 10 were specific to this state in comparison to the remission one (R-CRC) investigated on distinct patients. 15 metabolites involved in the correlations of interest, on the 25 unique ones obtained, were annotated (Metabolomics Standards Initiative level 2).

CONCLUSIONS

The metabolites highlighted could be used to better understand the pathology. The systematic investigation of the methodological aspects that we expose allows to implement correlation analysis to various fields and many specific cases.

A Lipid Signature with Perturbed Triacylglycerol Co-Regulation, Identified from Targeted Lipidomics, Predicts Risk for Type 2 Diabetes and Mediates the Risk from Adiposity in Two Prospective Cohorts of Chinese Adults

BACKGROUND

The roles of individual and co-regulated lipid molecular species in the development of type 2 diabetes (T2D) and mediation from metabolic risk factors remain unknown.

METHODS

We conducted profiling of 166 plasma lipid species in 2 nested case-control studies within 2 independent cohorts of Chinese adults, the Dongfeng-Tongji and the Jiangsu non-communicable disease cohorts. After 4.61 (0.15) and 7.57 (1.13) years' follow-up, 1039 and 520 eligible participants developed T2D in these 2 cohorts, respectively, and controls were 1:1 matched to cases by age and sex.

RESULTS

We found 27 lipid species, including 10 novel ones, consistently associated with T2D risk in the 2 cohorts. Differential correlation network analysis revealed significant correlations of triacylglycerol (TAG) 50:3, containing at least one oleyl chain, with 6 TAGs, at least 3 of which contain the palmitoyl chain, all downregulated within cases relative to controls among the 27 lipids in both cohorts, while the networks also both identified the oleyl chain-containing TAG 50:3 as the central hub. We further found that 13 of the 27 lipids consistently mediated the association between adiposity indicators (body mass index, waist circumference, and waist-to-height ratio) and diabetes risk in both cohorts (all P < 0.05; proportion mediated: 20.00%, 17.70%, and 17.71%, and 32.50%, 28.73%, and 33.86%, respectively).

CONCLUSIONS

Our findings suggested notable perturbed co-regulation, inferred from differential correlation networks, between oleyl chain- and palmitoyl chain-containing TAGs before diabetes onset, with the oleyl chain-containing TAG 50:3 at the center, and provided novel etiological insight regarding lipid dysregulation in the progression from adiposity to overt T2D.

SeqNet: An R Package for Generating Gene-Gene Networks and Simulating RNA-Seq Data

Gene expression data provide an abundant resource for inferring connections in gene regulatory networks. While methodologies developed for this task have shown success, a challenge remains in comparing the performance among methods. Gold-standard datasets are scarce and limited in use. And while tools for simulating expression data are available, they are not designed to resemble the data obtained from RNA-seq experiments. SeqNet is an R package that provides tools for generating a rich variety of gene network structures and simulating RNA-seq data from them. This produces in silico RNA-seq data for benchmarking and assessing gene network inference methods. The package is available on CRAN and on GitHub at https://github.com/tgrimes/SeqNet.

Characterizing Genetic Regulatory Elements in Ovine Tissues

The Ovine Functional Annotation of Animal Genomes (FAANG) project, part of the broader livestock species FAANG initiative, aims to identify and characterize gene regulatory elements in domestic sheep. Regulatory element annotation is essential for identifying genetic variants that affect health and production traits in this important agricultural species, as greater than 90% of variants underlying genetic effects are estimated to lie outside of transcribed regions. Histone modifications that distinguish active or repressed chromatin states, CTCF binding, and DNA methylation were used to characterize regulatory elements in liver, spleen, and cerebellum tissues from four yearling sheep. Chromatin immunoprecipitation with sequencing (ChIP-seq) was performed for H3K4me3, H3K27ac, H3K4me1, H3K27me3, and CTCF. Nine chromatin states including active promoters, active enhancers, poised enhancers, repressed enhancers, and insulators were characterized in each tissue using ChromHMM. Whole-genome bisulfite sequencing (WGBS) was performed to determine the complement of whole-genome DNA methylation with the ChIP-seq data. Hypermethylated and hypomethylated regions were identified across tissues, and these locations were compared with chromatin states to better distinguish and validate regulatory elements in these tissues. Interestingly, chromatin states with the poised enhancer mark H3K4me1 in the spleen and cerebellum and CTCF in the liver displayed the greatest number of hypermethylated sites. Not surprisingly, active enhancers in the liver and spleen, and promoters in the cerebellum, displayed the greatest number of hypomethylated sites. Overall, chromatin states defined by histone marks and CTCF occupied approximately 22% of the genome in all three tissues. Furthermore, the liver and spleen displayed in common the greatest percent of active promoter (65%) and active enhancer (81%) states, and the liver and cerebellum displayed in common the greatest percent of poised enhancer (53%), repressed enhancer (68%), hypermethylated sites (75%), and hypomethylated sites (73%). In addition, both known and de novo CTCF-binding motifs were identified in all three tissues, with the highest number of unique motifs identified in the cerebellum. In summary, this study has identified the regulatory regions of genes in three tissues that play key roles in defining health and economically important traits and has set the precedent for the characterization of regulatory elements in ovine tissues using the Rambouillet reference genome.

Comparing Statistical Tests for Differential Network Analysis of Gene Modules

Genes often work together to perform complex biological processes, and "networks" provide a versatile framework for representing the interactions between multiple genes. Differential network analysis (DiNA) quantifies how this network structure differs between two or more groups/phenotypes (e.g., disease subjects and healthy controls), with the goal of determining whether differences in network structure can help explain differences between phenotypes. In this paper, we focus on gene co-expression networks, although in principle, the methods studied can be used for DiNA for other types of features (e.g., metabolome, epigenome, microbiome, proteome, etc.). Three common applications of DiNA involve (1) testing whether the connections to a single gene differ between groups, (2) testing whether the connection between a pair of genes differs between groups, or (3) testing whether the connections within a "module" (a subset of 3 or more genes) differs between groups. This article focuses on the latter, as there is a lack of studies comparing statistical methods for identifying differentially co-expressed modules (DCMs). Through extensive simulations, we compare several previously proposed test statistics and a new p-norm difference test (PND). We demonstrate that the true positive rate of the proposed PND test is competitive with and often higher than the other methods, while controlling the false positive rate. The R package discoMod (differentially co-expressed modules) implements the proposed method and provides a full pipeline for identifying DCMs: clustering tools to derive gene modules, tests to identify DCMs, and methods for visualizing the results.

Metabolic networks of plasma and joint fluid base on differential correlation

Whether osteoarthritis (OA) is a systemic metabolic disorder remains controversial. The aim of this study was to investigate the metabolic characteristics between plasma and knee joint fluid (JF) of patients with advanced OA using a differential correlation metabolic (DCM) networks approach. Plasma and JF were collected during the joint replacement surgery of patients with knee OA. The biological samples were pretreated with standard procedures for metabolite analysis. The metabolic profiling was conducted by means of liquid mass spectrometry coupled with a AbsoluteIDQ kit. A DCM network approach was adopted for analyzing the metabolomics data between the plasma and JF. The variation in the correlation of the pairwise metabolites was quantified across the plasma and JF samples, and networks analysis was used to characterize the difference in the correlations of the metabolites from the two sample types. Core metabolites that played an important role in the DCM networks were identified via topological analysis. One hundred advanced OA patients (50 men and 50 women) were included in this study, with an average age of 65.0 ± 7.6 years (65.6 ± 7.1 years for females and 64.4 ± 8.1 years for males) and a mean BMI of 32.6 ± 5.8 kg/m² (33.4 ± 6.3 kg/m² for females and 31.7 ± 5.3 kg/m² for males). Age and BMI matched between the male and female groups. One hundred and forty-five nodes, 567 edges, and 131 nodes, 407 edges were found in the DCM networks (p < 0.05) of the female and male groups, respectively. Six metabolites in the female group and 5 metabolites in the male group were identified as key nodes in the network. There was a significant difference in the differential correlation metabolism networks of plasma and JF that may be related to local joint metabolism. Focusing on these key metabolites may help uncover the pathogenesis of knee OA. In addition, the differential metabolic correlation between plasma and JF mostly overlapped, indicating that these common correlations of pairwise metabolites may be a reflection of systemic characteristics of JF and that most significant correlation variations were just a result of "housekeeping” biological reactions.

MuscNet, a Weighted Voting Model of Multi-Source Connectivity Networks to Predict Mild Cognitive Impairment Using Resting-State Functional MRI

The neurological disorder mild cognitive impairment (MCI) demonstrates minor impacts on the patient's daily activities and may be ignored as the status of normal aging. But some of the MCI patients may further develop into severe statuses like Alzheimer's disease (AD). The brain functional connectivity network (BFCN) was usually constructed from the resting-state functional magnetic resonance imaging (rs-fMRI) data. This technology has been widely used to detect the neurodegenerative dementia and to reveal the intrinsic mechanism of neural activities. The BFCN edge was usually determined by the pairwise correlation between the brain regions. This study proposed a weighted voting model of multi-source connectivity networks (MuscNet) by integrating multiple BFCNs of different correlation coefficients. Our model was further improved by removing redundant features. The experimental data demonstrated that different BFCNs contributed complementary information to each other and MuscNet outperformed the existing models on detecting MCI patients. The previous study suggested the existence of multiple solutions with similarly good performance for a machine learning problem. The proposed model MuscNet utilized a weighted voting strategy to slightly outperform the existing studies, suggesting an effective way to fuse multiple base models. The reason may need further theoretical investigations about why different base models contribute to each other for the MCI prediction.

(Differential) Co-Expression Analysis of Gene Expression: A Survey of Best Practices

Analysis of gene expression data is widely used in transcriptomic studies to understand functions of molecules inside a cell and interactions among molecules. Differential co-expression analysis studies diseases and phenotypic variations by finding modules of genes whose co-expression patterns vary across conditions. We review the best practices in gene expression data analysis in terms of analysis of (differential) co-expression, co-expression network, differential networking, and differential connectivity considering both microarray and RNA-seq data along with comparisons. We highlight hurdles in RNA-seq data analysis using methods developed for microarrays. We include discussion of necessary tools for gene expression analysis throughout the paper. In addition, we shed light on scRNA-seq data analysis by including preprocessing and scRNA-seq in co-expression analysis along with useful tools specific to scRNA-seq. To get insights, biological interpretation and functional profiling is included. Finally, we provide guidelines for the analyst, along with research issues and challenges which should be addressed.

Classification of Cancer Types Using Graph Convolutional Neural Networks

BACKGROUND

Cancer has been a leading cause of death in the United States with significant health care costs. Accurate prediction of cancers at an early stage and understanding the genomic mechanisms that drive cancer development are vital to the improvement of treatment outcomes and survival rates, thus resulting in significant social and economic impacts. Attempts have been made to classify cancer types with machine learning techniques during the past two decades and deep learning approaches more recently.

RESULTS

In this paper, we established four models with graph convolutional neural network (GCNN) that use unstructured gene expressions as inputs to classify different tumor and non-tumor samples into their designated 33 cancer types or as normal. Four GCNN models based on a co-expression graph, co-expression+singleton graph, protein-protein interaction (PPI) graph, and PPI+singleton graph have been designed and implemented. They were trained and tested on combined 10,340 cancer samples and 731 normal tissue samples from The Cancer Genome Atlas (TCGA) dataset. The established GCNN models achieved excellent prediction accuracies (89.9-94.7%) among 34 classes (33 cancer types and a normal group). In silico gene-perturbation experiments were performed on four models based on co-expression graph, co-expression+singleton, PPI graph, and PPI+singleton graphs. The co-expression GCNN model was further interpreted to identify a total of 428 markers genes that drive the classification of 33 cancer types and normal. The concordance of differential expressions of these markers between the represented cancer type and others are confirmed. Successful classification of cancer types and a normal group regardless of normal tissues' origin suggested that the identified markers are cancer-specific rather than tissue-specific.

CONCLUSION

Novel GCNN models have been established to predict cancer types or normal tissue based on gene expression profiles. We demonstrated the results from the TCGA dataset that these models can produce accurate classification (above 94%), using cancer-specific markers genes. The models and the source codes are publicly available and can be readily adapted to the diagnosis of cancer and other diseases by the data-driven modeling research community.

Hippocampal Subregion Transcriptomic Profiles Reflect Strategy Selection during Cognitive Aging

Age-related cognitive impairments are associated with differentially expressed genes (DEGs) linked to defined neural systems; however, studies examining multiple regions of the hippocampus fail to find links between behavior and transcription in the dentate gyrus (DG). We hypothesized that use of a task requiring intact DG function would emphasize molecular signals in the DG associated with a decline in performance. We used a water maze beacon discrimination task to characterize young and middle-age male F344 rats, followed by a spatial reference memory probe trial test. Middle-age rats showed increased variability in discriminating two identical beacons. Use of an allocentric strategy and formation of a spatial reference memory were not different between age groups; however, older animals compensated for impaired beacon discrimination through greater reliance on spatial reference memory. mRNA sequencing of hippocampal subregions indicated DEGs in the DG of middle-age rats, linked to synaptic function and neurogenesis, correlated with beacon discrimination performance, suggesting that senescence of the DG underlies the impairment. Few genes correlated with spatial memory across age groups, with a greater number in region CA1. Age-related CA1 DEGs, correlated with spatial memory, were linked to regulation of neural activity. These results indicate that the beacon task is sensitive to impairment in middle age, and distinct gene profiles are observed in neural circuits that underlie beacon discrimination performance and allocentric memory. The use of different strategies in older animals and associated transcriptional profiles could provide an animal model for examining cognitive reserve and neural compensation of aging.SIGNIFICANCE STATEMENT Hippocampal subregions are thought to differentially contribute to memory. We took advantage of age-related variability in performance on a water maze beacon task and next-generation sequencing to test the hypothesis that aging of the dentate gyrus is linked to impaired beacon discrimination and compensatory use of allocentric memory. The dentate gyrus expressed synaptic function and neurogenesis genes correlated with beacon discrimination in middle-age animals. Spatial reference memory was associated with CA1 transcriptional correlates linked to regulation of neural activity and use of an allocentric strategy. This is the first study examining transcriptomes of multiple hippocampal subregions to link age-related impairments associated with discrimination of feature overlap and alternate response strategies to gene expression in specific hippocampal subregions.

DCARS: differential correlation across ranked samples

MOTIVATION

Genes act as a system and not in isolation. Thus, it is important to consider coordinated changes of gene expression rather than single genes when investigating biological phenomena such as the aetiology of cancer. We have developed an approach for quantifying how changes in the association between pairs of genes may inform the outcome of interest called Differential Correlation across Ranked Samples (DCARS). Modelling gene correlation across a continuous sample ranking does not require the dichotomisation of samples into two distinct classes and can identify differences in gene correlation across early, mid or late stages of the outcome of interest.

RESULTS

When we evaluated DCARS against the typical Fisher Z-transformation test for differential correlation, as well as a typical approach testing for interaction within a linear model, on real TCGA data, DCARS significantly ranked gene pairs containing known cancer genes more highly across several cancers. Similar results are found with our simulation study. DCARS was applied to 13 cancers datasets in TCGA, revealing several distinct relationships for which survival ranking was found to be associated with a change in correlation between genes. Furthermore, we demonstrated that DCARS can be used in conjunction with network analysis techniques to extract biological meaning from multi-layered and complex data.

AVAILABILITY AND IMPLEMENTATION

DCARS R package and sample data are available at https://github.com/shazanfar/DCARS. Publicly available data from The Cancer Genome Atlas (TCGA) was used using the TCGABiolinks R package. Supplementary Files and DCARS R package is available at https://github.com/shazanfar/DCARS.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Differential co-expression-based detection of conditional relationships in transcriptional data: comparative analysis and application to breast cancer

BACKGROUND

Elucidation of regulatory networks, including identification of regulatory mechanisms specific to a given biological context, is a key aim in systems biology. This has motivated the move from co-expression to differential co-expression analysis and numerous methods have been developed subsequently to address this task; however, evaluation of methods and interpretation of the resulting networks has been hindered by the lack of known context-specific regulatory interactions.

RESULTS

In this study, we develop a simulator based on dynamical systems modelling capable of simulating differential co-expression patterns. With the simulator and an evaluation framework, we benchmark and characterise the performance of inference methods. Defining three different levels of "true" networks for each simulation, we show that accurate inference of causation is difficult for all methods, compared to inference of associations. We show that a z-score-based method has the best general performance. Further, analysis of simulation parameters reveals five network and simulation properties that explained the performance of methods. The evaluation framework and inference methods used in this study are available in the dcanr R/Bioconductor package.

CONCLUSIONS

Our analysis of networks inferred from simulated data show that hub nodes are more likely to be differentially regulated targets than transcription factors. Based on this observation, we propose an interpretation of the inferred differential network that can reconstruct a putative causal network.

Laser Capture Microdissection and RNA-Seq Analysis: High Sensitivity Approaches to Explain Histopathological Heterogeneity in Human Glioblastoma FFPE Archived Tissues

Laser capture microdissection (LCM) coupled with RNA-seq is a powerful tool to identify genes that are differentially expressed in specific histological tumor subtypes. To better understand the role of single tumor cell populations in the complex heterogeneity of glioblastoma, we paired microdissection and NGS technology to study intra-tumoral differences into specific histological regions and cells of human GBM FFPE tumors. We here isolated astrocytes, neurons and endothelial cells in 6 different histological contexts: tumor core astrocytes, pseudopalisading astrocytes, perineuronal astrocytes in satellitosis, neurons with satellitosis, tumor blood vessels, and normal blood vessels. A customized protocol was developed for RNA amplification, library construction, and whole transcriptome analysis of each single portion. We first validated our protocol comparing the obtained RNA expression pattern with the gene expression levels of RNA-seq raw data experiments from the BioProject NCBI database, using Spearman's correlation coefficients calculation. We found a good concordance for pseudopalisading and tumor core astrocytes compartments (0.5 Spearman correlation) and a high concordance for perineuronal astrocytes, neurons, normal, and tumor endothelial cells compartments (0.7 Spearman correlation). Then, Principal Component Analysis and differential expression analysis were employed to find differences between tumor compartments and control tissue and between same cell types into distinct tumor contexts. Data consistent with the literature emerged, in which multiple therapeutic targets significant for glioblastoma (such as Integrins, Extracellular Matrix, transmembrane transport, and metabolic processes) play a fundamental role in the disease progression. Moreover, specific cellular processes have been associated with certain cellular subtypes within the tumor. Our results are promising and suggest a compelling method for studying glioblastoma heterogeneity in FFPE samples and its application in both prospective and retrospective studies.

New Statistical Methods for Constructing Robust Differential Correlation Networks to characterize the interactions among microRNAs

The interplay among microRNAs (miRNAs) plays an important role in the developments of complex human diseases. Co-expression networks can characterize the interactions among miRNAs. Differential correlation network is a powerful tool to investigate the differences of co-expression networks between cases and controls. To construct a differential correlation network, the Fisher's Z-transformation test is usually used. However, the Fisher's Z-transformation test requires the normality assumption, the violation of which would result in inflated Type I error rate. Several bootstrapping-based improvements for Fisher's Z test have been proposed. However, these methods are too computationally intensive to be used to construct differential correlation networks for high-throughput genomic data. In this article, we proposed six novel robust equal-correlation tests that are computationally efficient. The systematic simulation studies and a real microRNA data analysis showed that one of the six proposed tests (ST5) overall performed better than other methods.