Gene Coexpression Analyses Differentiate Networks Associated with Diverse Cancers Harboring TP53 Missense or Null Mutations

p53 missense mutation is associated with immune cell PD-L1 expression in triple-negative breast cancer

The programmed death ligand 1 (PD-L1) is a pivotal biomarker of immunotherapy in triple negative breast cancer (TNBC). TP53 is reported as a positive regulatory predictor of immune efficacy. The correlation of p53 expression or mutation and PD-L1 expression is explored. By immunohistochemistry, PD-L1 expression between p53 mutation (missense and nonsense) and wild type; p53 no-expression/loss vs. expression were compared. There was a significant association between p53 mutation, especially missense mutation with higher histological grade, and PD-L1 expression in immune cells (ICs). Both p53 missense mutation and PD-L1 expression may be potential targets for improving immunotherapy response in TNBC.

Aberrant calcium signalling downstream of mutations in TP53 and the PI3K/AKT pathway genes promotes disease progression and therapy resistance in triple negative breast cancer

Triple-negative breast cancer (TNBC) is characterized as an aggressive form of breast cancer (BC) associated with poor patient outcomes. For the majority of patients, there is a lack of approved targeted therapies. Therefore, chemotherapy remains a key treatment option for these patients, but significant issues around acquired resistance limit its efficacy. Thus, TNBC has an unmet need for new targeted personalized medicine approaches. Calcium (Ca²⁺) is a ubiquitous second messenger that is known to control a range of key cellular processes by mediating signalling transduction and gene transcription. Changes in Ca²⁺ through altered calcium channel expression or activity are known to promote tumorigenesis and treatment resistance in a range of cancers including BC. Emerging evidence shows that this is mediated by Ca²⁺ modulation, supporting the function of tumour suppressor genes (TSGs) and oncogenes. This review provides insight into the underlying alterations in calcium signalling and how it plays a key role in promoting disease progression and therapy resistance in TNBC which harbours mutations in tumour protein p53 (TP53) and the PI3K/AKT pathway.

Assessment of programmed death-ligand 1 receptor immunohistochemical expression and its association with tumor-infiltrating lymphocytes and p53 status in triple-negative breast cancer

Breast cancer (BC) is the second most frequent type of cancer for both sexes combined, after lung cancer. Triple-negative BC (TNBC) molecular subtype is characterized by lack of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) immunoexpression or amplification and represent 10-20% of all BC cases. The issue of the present study was to analyze the associations between programmed death-ligand 1 (PD-L1) immunoexpression and distribution of stromal tumor-infiltrating lymphocytes (stTILs) combined with clinico-morphological features of patients with TNBC. Secondly, our research evaluated PD-L1 immunoexpression as a prognostic factor and its correlation with p53 immunoexpression. Thirty cases with primary TNBC without prior neoadjuvant therapy were included in this research. stTILs were identified in all cases, most of them with low distribution (66.7%). A positive immunoreaction for PD-L1 was observed in 40% of cases. The PD-L1 immunoexpression was statistically significant associated with age, pathological tumor size, lymphovascular invasion, stTILs level, the presence of cluster of differentiation 8-positive (CD8+) TILs and p53 immunoexpression. In the present study, a positive PD-L1 immunoexpression was associated with a worse distant metastasis free survival (DMFS). We also found not only that high stTILs level were associated with a better DMFS but also that there was a statistically significant association between stTILs level and PD-L1 immunoexpression. Our results bring new insights to the fine connections between tumor microenvironment and molecular changes of TNBC. It helps us to better understand these aggressive tumors to identify the more useful biomarkers for predicting the response to adjuvant therapy and can represent a method for selecting the most suitable patients for immunotherapy.

Expression profile analysis to predict potential biomarkers for glaucoma: BMP1, DMD and GEM

Purpose

Glaucoma is the second commonest cause of blindness. We assessed the gene expression profile of astrocytes in the optic nerve head to identify possible prognostic biomarkers for glaucoma.

Method

A total of 20 patient and nine normal control subject samples were derived from the GSE9944 (six normal samples and 13 patient samples) and GSE2378 (three normal samples and seven patient samples) datasets, screened by microarray-tested optic nerve head tissues, were obtained from the Gene Expression Omnibus (GEO) database. We used a weighted gene coexpression network analysis (WGCNA) to identify coexpressed gene modules. We also performed a functional enrichment analysis and least absolute shrinkage and selection operator (LASSO) regression analysis. Genes expression was represented by boxplots, functional geneset enrichment analyses (GSEA) were used to profile the expression patterns of all the key genes. Then the key genes were validated by the external dataset.

Results

A total 8,606 genes and 19 human optic nerve head samples taken from glaucoma patients in the GSE9944 were compared with normal control samples to construct the co-expression gene modules. After selecting the most common clinical traits of glaucoma, their association with gene expression was established, which sorted two modules showing greatest correlations. One with the correlation coefficient is 0.56 (P = 0.01) and the other with the correlation coefficient is −0.56 (P = 0.01). Hub genes of these modules were identified using scatterplots of gene significance versus module membership. A functional enrichment analysis showed that the former module was mainly enriched in genes involved in cellular inflammation and injury, whereas the latter was mainly enriched in genes involved in tissue homeostasis and physiological processes. This suggests that genes in the green–yellow module may play critical roles in the onset and development of glaucoma. A LASSO regression analysis identified three hub genes: Recombinant Bone Morphogenetic Protein 1 gene (BMP1), Duchenne muscular dystrophy gene (DMD) and mitogens induced GTP-binding protein gene (GEM). The expression levels of the three genes in the glaucoma group were significantly lower than those in the normal group. GSEA further illuminated that BMP1, DMD and GEM participated in the occurrence and development of some important metabolic progresses. Using the GSE2378 dataset, we confirmed the high validity of the model, with an area under the receiver operator characteristic curve of 85%.

Conclusion

We identified several key genes, including BMP1, DMD and GEM, that may be involved in the pathogenesis of glaucoma. Our results may help to determine the prognosis of glaucoma and/or to design gene- or molecule-targeted drugs.

Meta-Analysis Based on Nonconvex Regularization

The widespread applications of high-throughput sequencing technology have produced a large number of publicly available gene expression datasets. However, due to the gene expression datasets have the characteristics of small sample size, high dimensionality and high noise, the application of biostatistics and machine learning methods to analyze gene expression data is a challenging task, such as the low reproducibility of important biomarkers in different studies. Meta-analysis is an effective approach to deal with these problems, but the current methods have some limitations. In this paper, we propose the meta-analysis based on three nonconvex regularization methods, which are L1/2 regularization (meta-Half), Minimax Concave Penalty regularization (meta-MCP) and Smoothly Clipped Absolute Deviation regularization (meta-SCAD). The three nonconvex regularization methods are effective approaches for variable selection developed in recent years. Through the hierarchical decomposition of coefficients, our methods not only maintain the flexibility of variable selection and improve the efficiency of selecting important biomarkers, but also summarize and synthesize scientific evidence from multiple studies to consider the relationship between different datasets. We give the efficient algorithms and the theoretical property for our methods. Furthermore, we apply our methods to the simulation data and three publicly available lung cancer gene expression datasets, and compare the performance with state-of-the-art methods. Our methods have good performance in simulation studies, and the analysis results on the three publicly available lung cancer gene expression datasets are clinically meaningful. Our methods can also be extended to other areas where datasets are heterogeneous.

Identification of differentially expressed genes between the colon and ileum of patients with inflammatory bowel disease by gene co-expression analysis

OBJECTIVE

We aimed to identify differentially expressed genes (DEG) in patients with inflammatory bowel disease (IBD).

METHODS

RNA-seq data were obtained from the Array Express database. DEG were identified using the edgeR package. A co-expression network was constructed and key modules with the highest correlation with IBD inflammatory sites were identified for analysis. The Cytoscape MCODE plugin was used to identify key sub-modules of the protein-protein interaction (PPI) network. The genes in the sub-modules were considered hub genes, and functional enrichment analysis was performed. Furthermore, we constructed a drug-gene interaction network. Finally, we visualized the hub gene expression pattern between the colon and ileum of IBD using the ggpubr package and analyzed it using the Wilcoxon test.

RESULTS

DEG were identified between the colon and ileum of IBD patients. Based on the co-expression network, the green module had the highest correlation with IBD inflammatory sites. In total, 379 DEG in the green module were identified for the PPI network. Nineteen hub genes were differentially expressed between the colon and ileum. The drug-gene network identified these hub genes as potential drug targets.

CONCLUSION

Nineteen DEG were identified between the colon and ileum of IBD patients.

A Six-Gene Signature Predicts Survival of Adenocarcinoma Type of Non-Small-Cell Lung Cancer Patients: A Comprehensive Study Based on Integrated Analysis and Weighted Gene Coexpression Network

Background and Goals. To identify a multigene signature model for prognosis of non-small-cell lung cancer (NSCLC) patients, we first found 2146 consensus differentially expressed genes (DEGs) in NSCLC overlapped in Gene Expression Omnibus (GEO) and TCGA lung adenocarcinoma (LUAD) datasets using integrated analysis. We constructed a weighted gene coexpression network (WGCN) using the consensus DEGs and identified the module significantly associated with pathological M stage and consisted of 61 genes. After univariate Cox regression analysis and subsequent stepwise model selection by the Akaike information criterion (AIC) and multivariate Cox hazard model analysis, an mRNA signature model which calculated prognostic score was generated: prognostic score = (-0.2491 × EXP_RRAGB) + (-0.0679 × EXP_RSPH9) + (-0.2317 × EXP_RPS6KL1) + (-0.1035 × EXP_RXFP1) + 0.1571 × EXP_RRM2 + 0.1104 × EXP_RTL1, where EXP is the fragments per kilobase million (FPKM) value of the mRNA included in the model. The prognostic model separated NSCLC patients in the TCGA-LUAD dataset into the low- and high-risk score groups with a median prognostic score of 0.972. Higher scores predicted higher risk. The area under ROC curve (AUC) was 0.994 or 0.776 in predicting the 1- to 10-year survival of NSCLC patients. The prognostic performance of this prognostic model was validated by an independent GSE11969 dataset of NSCLC adenocarcinoma with AUC values between 0.822 and 0.755 in predicting 1- to 10-year survival of NSCLC. These results suggested that the six-gene signature functioned as an independent biomarker to predict the overall survival of NSCLC adenocarcinoma.

Gene networks in cancer are biased by aneuploidies and sample impurities

Gene regulatory network inference is a standard technique for obtaining structured regulatory information from, for instance, gene expression measurements. Methods performing this task have been extensively evaluated on synthetic, and to a lesser extent real data sets. In contrast to these test evaluations, applications to gene expression data of human cancers are often limited by fewer samples and more potential regulatory links, and are biased by copy number aberrations as well as cell mixtures and sample impurities. Here, we take networks inferred from TCGA cohorts as an example to show that (1) transcription factor annotations are essential to obtain reliable networks, and (2) even for state of the art methods, we expect that between 20 and 80% of edges are caused by copy number changes and cell mixtures rather than transcription factor regulation.

Identifying Potential Prognostic Markers for Muscle-Invasive Bladder Urothelial Carcinoma by Weighted Gene Co-Expression Network Analysis

Muscle-invasive bladder urothelial carcinoma (MIBC) is characterized as a genetic heterogeneous cancer with a high percentage of recurrence and worse prognosis. Identify the prognostic potentials of novel genes for muscle-invasive urothelial bladder cancer could at least provide important information for early detection and clinical treatment. Weighted gene co-expression network analysis (WGCNA) algorithm, a powerful systems biology approach, was utilized to extract co-expressed gene networks from mRNA expression dataset to construct transcriptional modules in MIBC samples, which was associated with demographic and clinical traits of MIBC patients. The potential prognostic markers of MIBC were screened out in the discovery dataset and verified in an independent external validation dataset. A total of 8 co-expression modules were detected through the WGCNA algorithm in the discovery datasets based on 401 MIBC samples. One transcriptional module enriched in cell development was observed to be correlated with the MIBC prognosis in the discovery datasets (HR = 1.48, 95%CI = 1.04-2.11) and independently verified in an external dataset (HR = 3.59, 95%CI = 1.09-11.79). High expression of hub genes including discoidin domain receptor tyrosine kinase 2 (DDR2), PDZ and LIM domain 3 (PDLIM3), zinc finger protein 521 (ZNF521), methionine sulfoxide reductase B3 (MSRB3) were significantly associated with the unfavorable survival of MIBC patients. We identified and validated four novel potential biomarkers associated with prognosis of MIBC patients by constructing genes co-expression networks. The discovery of these genetic markers may provide a new target for the development of MIBC chemotherapeutic drugs.

Association between p53 Expression and Amount of Tumor-Infiltrating Lymphocytes in Triple-Negative Breast Cancer

Background

Most triple-negative breast cancers (TNBCs) have a high histologic grade, are associated with high endoplasmic stress, and possess a high frequency of TP53 mutations. TP53 missense mutations lead to the production of mutant p53 protein and usually show high levels of p53 protein expression. Tumor-infiltrating lymphocytes (TILs) accumulate as part of the anti-tumor immune response and have a strong prognostic and predictive significance in TNBC. We aimed to elucidate the association between p53 expression and the amount of TILs in TNBC.

Methods

In 678 TNBC patients, we evaluated TIL levels and expression of endoplasmic stress molecules. Immunohistochemical examination of p53 protein expression was categorized into three groups: no, low, and high expression.

Results

No, low, and high p53 expression was identified in 44.1% (n = 299), 20.1% (n = 136), and 35.8% (n = 243) of patients, respectively. Patients with high p53 expression showed high histologic grade (p < .001), high TIL levels (p = .009), and high expression of endoplasmic reticulum stress-associated molecules (p-eIF2a, p = .013; XBP1, p = .007), compared to patients with low p53 expression. There was no significant difference in disease-free (p = .406) or overall survival rates (p = .444) among the three p53 expression groups.

Conclusions

High p53 expression is associated with increased expression of endoplasmic reticulum stress molecules and TIL influx.

Identification of lncRNA TRPM2-AS/miR-140-3p/PYCR1 axis's proliferates and anti-apoptotic effect on breast cancer using co-expression network analysis

Breast cancer (BC) is one of the most common malignancies occurring in women worldwide. Weighted gene co-expression network analysis (WGCNA) has not been widely utilized in uncovering the biomarkers which played pivotal roles in BC treatment. This study aimed to verify the proliferative and anti-apoptotic effect of lncRNA TRPM2-AS/miR-140-3p/PYCR1 axis on BC based on WGCNA. WGCNA was applied for determining hub genes using gene expression data gained from breast cancer and adjacent tissues which were downloaded from the Cancer Genome Atlas (TCGA) database. The correlative curves showed the correlation between OS/DFS of BC patients and TRPM2-AS expression or PYCR1 expression based on the data of survival rate of BC patients obtained from the TCGA database. QRT-PCR was employed in detecting the expression levels of TRPM2-AS, miR-140-3p and PYCR1, and western blot analysis was adopted for determination of protein expression level of PYCR1. Dual luciferase assay was applied to verify the targeting relationship between TRPM2-AS and miR-140-3p, as well as miR-140-3p and PYCR1. The roles of TRPM2-AS, miR-140-3p, and PYCR1 in proliferation, migration, and apoptosis of BC cell were identified by CCK-8 assay, cell migration assay and flow cytometry. Hub genes were also gained from WGCNA test. The prognostic study showed a significant negative correlation between the high expression of PYCR1 and TRPM2-AS and the BC survival. QRT-PCR demonstrated that PYCR1 and TRPM2-AS were both overexpressed, while miR-140-3p was greatly down-regulated in BC cell. In addition, it was validated by dual luciferase assay that miR-140-3p directly targeted both TRPM2-AS and PYCR1. Furthermore, down-regulation of TRPM2-AS and PYCR1 inhibited proliferation yet promoted apoptosis of BC cell, and up-regulation of miR-140-3p in BC cell showed the same tendency. Taken together, TRPM2-AS could promote proliferation and inhibit apoptosis of BC cell through TRPM2-AS/miR-140-3p/PYCR1 axis.

Differential Coexpression Network Analysis for Gene Expression Data

Gene expression profiling by microarray has been used to uncover molecular variations in many areas. The traditional analysis method to gene expression profiling just focuses on the individual genes, and the interactions among genes are ignored, while genes play their roles not by isolations but by interactions with each other. Consequently, gene-to-gene coexpression analysis emerged as a powerful approach to solve the above problems. Then complementary to the conventional differential expression analysis, the differential coexpression analysis can identify gene markers from the systematic level. There are three aspects for differential coexpression network analysis including the network global topological comparison, differential coexpression module identification, and differential coexpression genes and gene pairs identification. To date, the coexpression network and differential coexpression analysis are widely used in a variety of areas in response to environmental stresses, genetic differences, or disease changes. In this chapter, we reviewed the existing methods for differential coexpression network analysis and discussed the applications to cancer research.

Network Rewiring in Cancer: Applications to Melanoma Cell Lines and the Cancer Genome Atlas Patients

Genes do not work in isolation, but rather as part of networks that have many feedback and redundancy mechanisms. Studying the properties of genetic networks and how individual genes contribute to overall network functions can provide insight into genetically-mediated disease processes. Most analytical techniques assume a network topology based on normal state networks. However, gene perturbations often lead to the rewiring of relevant networks and impact relationships among other genes. We apply a suite of analysis methodologies to assess the degree of transcriptional network rewiring observed in different sets of melanoma cell lines using whole genome gene expression microarray profiles. We assess evidence for network rewiring in melanoma patient tumor samples using RNA-sequence data available from The Cancer Genome Atlas. We make a distinction between “unsupervised” and “supervised” network-based methods and contrast their use in identifying consistent differences in networks between subsets of cell lines and tumor samples. We find that different genes play more central roles within subsets of genes within a broader network and hence are likely to be better drug targets in a disease state. Ultimately, we argue that our results have important implications for understanding the molecular pathology of melanoma as well as the choice of treatments to combat that pathology.

Gene co-expression network reveals shared modules predictive of stage and grade in serous ovarian cancers

Serous ovarian cancer (SOC) is the most lethal gynecological cancer. Clinical studies have revealed an association between tumor stage and grade and clinical prognosis. Identification of meaningful clusters of co-expressed genes or representative biomarkers related to stage or grade may help to reveal mechanisms of tumorigenesis and cancer development, and aid in predicting SOC patient prognosis. We therefore performed a weighted gene co-expression network analysis (WGCNA) and calculated module-trait correlations based on three public microarray datasets (GSE26193, GSE9891, and TCGA), which included 788 samples and 10402 genes. We detected four modules related to one or more clinical features significantly shared across all modeling datasets, and identified one stage-associated module and one grade-associated module. Our analysis showed that MMP2, COL3A1, COL1A2, FBN1, COL5A1, COL5A2, and AEBP1 are top hub genes related to stage, while CDK1, BUB1, BUB1B, BIRC5, AURKB, CENPA, and CDC20 are top hub genes related to grade. Gene and pathway enrichment analyses of the regulatory networks involving hub genes suggest that extracellular matrix interactions and mitotic signaling pathways are crucial determinants of tumor stage and grade. The relationships between gene expression modules and tumor stage or grade were validated in five independent datasets. These results could potentially be developed into a more objective scoring system to improve prediction of SOC outcomes.

Identifying biomarkers of papillary renal cell carcinoma associated with pathological stage by weighted gene co-expression network analysis

Although papillary renal cell carcinoma (PRCC) accounts for 10%–15% of renal cell carcinoma (RCC), no predictive molecular biomarker is currently applicable to guiding disease stage of PRCC patients. The mRNASeq data of PRCC and adjacent normal tissue in The Cancer Genome Atlas was analyzed to identify 1148 differentially expressed genes, on which weighted gene co-expression network analysis was performed. Then 11 co-expressed gene modules were identified. The highest association was found between blue module and pathological stage (r = 0.45) by Pearson's correlation analysis. Functional enrichment analysis revealed that biological processes of blue module focused on nuclear division, cell cycle phase, and spindle (all P < 1e-10). All 40 hub genes in blue module can distinguish localized (pathological stage I, II) from non-localized (pathological stage III, IV) PRCC (P < 0.01). A good molecular biomarker for pathological stage of RCC must be a prognostic gene in clinical practice. Survival analysis was performed to reversely validate if hub genes were associated with pathological stage. Survival analysis unveiled that all hub genes were associated with patient prognosis (P < 0.01). The validation cohort GSE2748 verified that 30 hub genes can differentiate localized from non-localized PRCC (P < 0.01), and 18 hub genes are prognosis-associated (P < 0.01).

ROC curve indicated that the 17 hub genes exhibited excellent diagnostic efficiency for localized and non-localized PRCC (AUC > 0.7). These hub genes may serve as a biomarker and help to distinguish different pathological stages for PRCC patients.

An analytic approach for interpretable predictive models in high-dimensional data in the presence of interactions with exposures

Predicting a phenotype and understanding which variables improve that prediction are two very challenging and overlapping problems in the analysis of high-dimensional (HD) data such as those arising from genomic and brain imaging studies. It is often believed that the number of truly important predictors is small relative to the total number of variables, making computational approaches to variable selection and dimension reduction extremely important. To reduce dimensionality, commonly used two-step methods first cluster the data in some way, and build models using cluster summaries to predict the phenotype. It is known that important exposure variables can alter correlation patterns between clusters of HD variables, that is, alter network properties of the variables. However, it is not well understood whether such altered clustering is informative in prediction. Here, assuming there is a binary exposure with such network-altering effects, we explore whether the use of exposure-dependent clustering relationships in dimension reduction can improve predictive modeling in a two-step framework. Hence, we propose a modeling framework called ECLUST to test this hypothesis, and evaluate its performance through extensive simulations. With ECLUST, we found improved prediction and variable selection performance compared to methods that do not consider the environment in the clustering step, or to methods that use the original data as features. We further illustrate this modeling framework through the analysis of three data sets from very different fields, each with HD data, a binary exposure, and a phenotype of interest. Our method is available in the eclust CRAN package.

Differential gene regulatory networks in development and disease

Gene regulatory networks, in which differential expression of regulator genes induce differential expression of their target genes, underlie diverse biological processes such as embryonic development, organ formation and disease pathogenesis. An archetypical systems biology approach to mapping these networks involves the combined application of (1) high-throughput sequencing-based transcriptome profiling (RNA-seq) of biopsies under diverse network perturbations and (2) network inference based on gene-gene expression correlation analysis. The comparative analysis of such correlation networks across cell types or states, differential correlation network analysis, can identify specific molecular signatures and functional modules that underlie the state transition or have context-specific function. Here, we review the basic concepts of network biology and correlation network inference, and the prevailing methods for differential analysis of correlation networks. We discuss applications of gene expression network analysis in the context of embryonic development, cancer, and congenital diseases.

Identification of two novel biomarkers of rectal carcinoma progression and prognosis via co-expression network analysis

mRNA expression profiles provide important insights on a diversity of biological processes involved in rectal carcinoma (RC). Our aim was to comprehensively map complex interactions between the mRNA expression patterns and the clinical traits of RC. We employed the integrated analysis of five microarray datasets and The Cancer Genome Atlas rectal adenocarcinoma database to identify 2118 consensual differentially expressed genes (DEGs) in RC and adjacent normal tissue samples, and then applied weighted gene co-expression network analysis to parse DEGs and eight clinical traits in 66 eligible RC samples. A total of 16 co-expressed gene modules were identified. The green-yellow and salmon modules were most appropriate to the pathological stage (R = 0.36) and the overall survival (HR =13.534, P = 0.014), respectively. A diagnostic model of the five pathological stage hub genes (SCG3, SYP, CDK5R2, AP3B2, and RUNDC3A) provided a powerful classification accuracy between localized RC and non-localized RC. We also found increased Secretogranin III (SCG3) expression with higher pathological stage and poorer prognosis in the test and validation set. The increased Homer scaffolding protein 2 (HOMER2) expression with the favorable survival prediction efficiency significantly correlated with the markedly reduced overall survival of RC patients and the higher pathological stage during the test and validation set. Our findings indicate that the SCG3 and HOMER2 mRNA levels should be further evaluated as predictors of pathological stage and survival in patients with RC.