ROS-DET: robust detector of switching mechanisms in gene expression

Serum lipidome reveals lipid metabolic dysregulation in severe fever with thrombocytopenia syndrome

BACKGROUND

Severe fever with thrombocytopenia syndrome (SFTS) is a rapidly progressing infectious disease with a high fatality rate caused by a novel bunyavirus (SFTSV). The role of lipids in viral infections is well-documented; however, the specific alterations in lipid metabolism during SFTSV infection remain elusive. This study aims to elucidate the lipid metabolic dysregulations in the early stages of SFTS patients.

METHODS

This study prospectively collected peripheral blood sera from 11 critical SFTS patients, 37 mild SFTS patients, and 23 healthy controls during the early stages of infection for lipidomics analysis. A systematic bioinformatics analysis was conducted from three aspects integrating lipid differential expressions, lipid differential correlations, and lipid-clinical indices correlations to reveal the serum lipid metabolic dysregulation in SFTSV-infected individuals.

RESULTS

Our findings reveal significant lipid metabolic dysregulation in SFTS patients. Specifically, compared to healthy controls, SFTS patients exhibited three distinct modes of lipid differential expression: increased levels of lipids including phosphatidylserine (PS), hexosylceramide (HexCer), and triglycerides (TG); decreased levels of lipids including lysophosphatidylcholine (LPC), acylcarnitine (AcCa), and cholesterol esters (ChE); and lipids showing "dual changes" including phosphatidylcholine (PC) and phosphatidylethanolamine (PE). Finally, based on lipid metabolic pathways and literature analysis, we systematically elucidated the potential mechanisms underlying lipid metabolic dysregulation in the early stage of SFTSV infection.

CONCLUSIONS

Our study presents the first global serum lipidome profile and reveals the lipid metabolic dysregulation patterns in the early stage of SFTSV infection. These findings provide a new basis for the diagnosis, treatment, and further investigation of the disease.

Identification and Validation of Key Genes of Differential Correlations in Gastric Cancer

Background: Gastric cancer (GC) is aggressive cancer with a poor prognosis. Previously bulk transcriptome analysis was utilized to identify key genes correlated with the development, progression and prognosis of GC. However, due to the complexity of the genetic mutations, there is still an urgent need to recognize core genes in the regulatory network of GC.

Methods: Gene expression profiles (GSE66229) were retrieved from the GEO database. Weighted correlation network analysis (WGCNA) was employed to identify gene modules mostly correlated with GC carcinogenesis. R package ‘DiffCorr’ was applied to identify differentially correlated gene pairs in tumor and normal tissues. Cytoscape was adopted to construct and visualize the gene regulatory network.

Results: A total of 15 modules were detected in WGCNA analysis, among which three modules were significantly correlated with GC. Then genes in these modules were analyzed separately by “DiffCorr”. Multiple differentially correlated gene pairs were recognized and the network was visualized by the software Cytoscape. Moreover, GEMIN5 and PFDN2, which were rarely discussed in GC, were identified as key genes in the regulatory network and the differential expression was validated by real-time qPCR, WB and IHC in cell lines and GC patient tissues.

Conclusions: Our research has shed light on the carcinogenesis mechanism by revealing differentially correlated gene pairs during transition from normal to tumor. We believe the application of this network-based algorithm holds great potential in inferring relationships and detecting candidate biomarkers.

Identification of Key Genes With Differential Correlations in Lung Adenocarcinoma

Background

With the advent of large-scale molecular profiling, an increasing number of oncogenic drivers contributing to precise medicine and reshaping classification of lung adenocarcinoma (LUAD) have been identified. However, only a minority of patients archived improved outcome under current standard therapies because of the dynamic mutational spectrum, which required expanding susceptible gene libraries. Accumulating evidence has witnessed that understanding gene regulatory networks as well as their changing processes was helpful in identifying core genes which acted as master regulators during carcinogenesis. The present study aimed at identifying key genes with differential correlations between normal and tumor status.

Methods

Weighted gene co-expression network analysis (WGCNA) was employed to build a gene interaction network using the expression profile of LUAD from The Cancer Genome Atlas (TCGA). R package DiffCorr was implemented for the identification of differential correlations between tumor and adjacent normal tissues. STRING and Cytoscape were used for the construction and visualization of biological networks.

Results

A total of 176 modules were detected in the network, among which yellow and medium orchid modules showed the most significant associations with LUAD. Then genes in these two modules were further chosen to evaluate their differential correlations. Finally, dozens of novel genes with opposite correlations including ATP13A4-AS1, HIGD1B, DAP3, and ISG20L2 were identified. Further biological and survival analyses highlighted their potential values in the diagnosis and treatment of LUAD. Moreover, real-time qPCR confirmed the expression patterns of ATP13A4-AS1, HIGD1B, DAP3, and ISG20L2 in LUAD tissues and cell lines.

Conclusion

Our study provided new insights into the gene regulatory mechanisms during transition from normal to tumor, pioneering a network-based algorithm in the application of tumor etiology.

(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.

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.

Three-way interaction model with switching mechanism as an effective strategy for tracing functionally-related genes

Introduction: Identification of functionally-related genes is an important step in understanding biological systems. The most popular strategy to infer functional dependence is to study pairwise correlations between gene expression levels. However, certain functionally-related genes may have a low expression correlation due to their nonlinear interactions. The use of a three-way interaction (3WI) model with switching mechanism (SM) is a relatively new strategy to trace functionally-related genes. The 3WI model traces the dynamic and nonlinear nature of the co-expression relationship of two genes by introducing their link to the expression level of a third gene. Areas covered: In this paper, we reviewed a variety of existing methods for tracing the 3WIs. Furthermore, we provide a comprehensive review of the previous biological studies based on 3WI models. Expert commentary: Comparison of features of these methods indicates that the modified liquid association algorithm has the best efficiency for tracing 3WI between others. The limited number of biological studies based on the 3WI suggests that high computational demand of the available algorithms is a major challenge to apply this approach for analyzing high-throughput omics data.

Differential correlation for sequencing data

BACKGROUND

Several methods have been developed to identify differential correlation (DC) between pairs of molecular features from -omics studies. Most DC methods have only been tested with microarrays and other platforms producing continuous and Gaussian-like data. Sequencing data is in the form of counts, often modeled with a negative binomial distribution making it difficult to apply standard correlation metrics. We have developed an R package for identifying DC called Discordant which uses mixture models for correlations between features and the Expectation Maximization (EM) algorithm for fitting parameters of the mixture model. Several correlation metrics for sequencing data are provided and tested using simulations. Other extensions in the Discordant package include additional modeling for different types of differential correlation, and faster implementation, using a subsampling routine to reduce run-time and address the assumption of independence between molecular feature pairs.

RESULTS

With simulations and breast cancer miRNA-Seq and RNA-Seq data, we find that Spearman's correlation has the best performance among the tested correlation methods for identifying differential correlation. Application of Spearman's correlation in the Discordant method demonstrated the most power in ROC curves and sensitivity/specificity plots, and improved ability to identify experimentally validated breast cancer miRNA. We also considered including additional types of differential correlation, which showed a slight reduction in power due to the additional parameters that need to be estimated, but more versatility in applications. Finally, subsampling within the EM algorithm considerably decreased run-time with negligible effect on performance.

CONCLUSIONS

A new method and R package called Discordant is presented for identifying differential correlation with sequencing data. Based on comparisons with different correlation metrics, this study suggests Spearman's correlation is appropriate for sequencing data, but other correlation metrics are available to the user depending on the application and data type. The Discordant method can also be extended to investigate additional DC types and subsampling with the EM algorithm is now available for reduced run-time. These extensions to the R package make Discordant more robust and versatile for multiple -omics studies.

Plasma microRNA biomarker detection for mild cognitive impairment using differential correlation analysis

BACKGROUND

Mild cognitive impairment (MCI) is an intermediate state between normal aging and dementia including Alzheimer's disease. Early detection of dementia, and MCI, is a crucial issue in terms of secondary prevention. Blood biomarker detection is a possible way for early detection of MCI. Although disease biomarkers are detected by, in general, using single molecular analysis such as t-test, another possible approach is based on interaction between molecules.

RESULTS

Differential correlation analysis, which detects difference on correlation of two variables in case/control study, was carried out to plasma microRNA (miRNA) expression profiles of 30 age- and race-matched controls and 23 Japanese MCI patients. The 20 pairs of miRNAs, which consist of 20 miRNAs, were selected as MCI markers. Two pairs of miRNAs (hsa-miR-191 and hsa-miR-101, and hsa-miR-103 and hsa-miR-222) out of 20 attained the highest area under the curve (AUC) value of 0.962 for MCI detection. Other two miRNA pairs that include hsa-miR-191 and hsa-miR-125b also attained high AUC value of ≥ 0.95. Pathway analysis was performed to the MCI markers for further understanding of biological implications. As a result, collapsed correlation on hsa-miR-191 and emerged correlation on hsa-miR-125b might have key role in MCI and dementia progression.

CONCLUSION

Differential correlation analysis, a bioinformatics tool to elucidate complicated and interdependent biological systems behind diseases, detects effective MCI markers that cannot be found by single molecule analysis such as t-test.

BFDCA: A Comprehensive Tool of Using Bayes Factor for Differential Co-Expression Analysis

Comparing the gene-expression profiles between biological conditions is useful for understanding gene regulation underlying complex phenotypes. Along this line, analysis of differential co-expression (DC) has gained attention in the recent years, where genes under one condition have different co-expression patterns compared with another. We developed an R package Bayes Factor approach for Differential Co-expression Analysis (BFDCA) for DC analysis. BFDCA is unique in integrating various aspects of DC patterns (including Shift, Cross, and Re-wiring) into one uniform Bayes factor. We tested BFDCA using simulation data and experimental data. Simulation results indicate that BFDCA outperforms existing methods in accuracy and robustness of detecting DC pairs and DC modules. Results of using experimental data suggest that BFDCA can cluster disease-related genes into functional DC subunits and estimate the regulatory impact of disease-related genes well. BFDCA also achieves high accuracy in predicting case-control phenotypes by using significant DC gene pairs as markers. BFDCA is publicly available at http://dx.doi.org/10.17632/jdz4vtvnm3.1.

Detecting Differentially Coexpressed Genes from Labeled Expression Data: A Brief Review

We review methods for capturing differential coexpression, which can be divided into two cases by the size of gene sets: 1) two paired genes and 2) multiple genes. In the first case, two genes are positively and negatively correlated with each other under one and the other conditions, respectively. In the second case, multiple genes are coexpressed and randomly expressed under one and the other conditions, respectively. We summarize a variety of methods for the first and second cases into four and three approaches, respectively. We describe each of these approaches in detail technically, being followed by thorough comparative experiments with both synthetic and real data sets. Our experimental results imply high possibility of improving the efficiency of the current methods, particularly in the case of multiple genes, because of low performance achieved by the best methods which are relatively simple intuitive ones.

DCGL v2.0: an R package for unveiling differential regulation from differential co-expression

Motivation

Differential co-expression analysis (DCEA) has emerged in recent years as a novel, systematic investigation into gene expression data. While most DCEA studies or tools focus on the co-expression relationships among genes, some are developing a potentially more promising research domain, differential regulation analysis (DRA). In our previously proposed R package DCGL v1.0, we provided functions to facilitate basic differential co-expression analyses; however, the output from DCGL v1.0 could not be translated into differential regulation mechanisms in a straightforward manner.

Results

To advance from DCEA to DRA, we upgraded the DCGL package from v1.0 to v2.0. A new module named “Differential Regulation Analysis” (DRA) was designed, which consists of three major functions: DRsort, DRplot, and DRrank. DRsort selects differentially regulated genes (DRGs) and differentially regulated links (DRLs) according to the transcription factor (TF)-to-target information. DRrank prioritizes the TFs in terms of their potential relevance to the phenotype of interest. DRplot graphically visualizes differentially co-expressed links (DCLs) and/or TF-to-target links in a network context. In addition to these new modules, we streamlined the codes from v1.0. The evaluation results proved that our differential regulation analysis is able to capture the regulators relevant to the biological subject.

Conclusions

With ample functions to facilitate differential regulation analysis, DCGL v2.0 was upgraded from a DCEA tool to a DRA tool, which may unveil the underlying differential regulation from the observed differential co-expression. DCGL v2.0 can be applied to a wide range of gene expression data in order to systematically identify novel regulators that have not yet been documented as critical.

Availability

DCGL v2.0 package is available at http://cran.r-project.org/web/packages/DCGL/index.html or at our project home page http://lifecenter.sgst.cn/main/en/dcgl.jsp.

DiffCorr: an R package to analyze and visualize differential correlations in biological networks

Large-scale "omics" data, such as microarrays, can be used to infer underlying cellular regulatory networks in organisms, enabling us to better understand the molecular basis of disease and important traits. Correlation approaches, such as a hierarchical cluster analysis, have been widely used to analyze omics data. In addition to the changes in the mean levels of molecules in the omics data, it is important to know about the changes in the correlation relationship among molecules between 2 experimental conditions. The development of a tool to identify differential correlation patterns in omics data in an efficient and unbiased manner is therefore desirable. We developed the DiffCorr package, a simple method for identifying pattern changes between 2 experimental conditions in correlation networks, which builds on a commonly used association measure, such as Pearson's correlation coefficient. DiffCorr calculates correlation matrices for each dataset, identifies the first principal component-based "eigen-molecules" in the correlation networks, and tests differential correlation between the 2 groups based on Fisher's z-test. We illustrated its utility by demonstrating biologically relevant, differentially correlated molecules in transcriptome coexpression and metabolite-to-metabolite correlation networks. DiffCorr can explore differential correlations between 2 conditions in the context of post-genomics data types, namely transcriptomics and metabolomics. DiffCorr is simple to use in calculating differential correlations and is suitable for the first step towards inferring causal relationships and detecting biomarker candidates. The package can be downloaded from the following website: http://diffcorr.sourceforge.net/.