Discovering Condition-Specific Gene Co-Expression Patterns Using Gaussian Mixture Models: A Cancer Case Study

Machine learning developed a PI3K/Akt pathway-related signature for predicting prognosis and drug sensitivity in ovarian cancer

BACKGROUND

Ovarian cancer is one of the deadliest malignancies among females, generally having a poor prognosis. The PI3K/Akt pathway plays a vital role in the oncogenesis and progression of many types of cancer. Limited studies have fully clarified the role of PI3K/Akt pathway in the prognosis of ovarian cancer and its correlation with drug sensitivity.

METHODS

A prognostic PI3K/Akt pathway related signature (PRS) was constructed with 10 machine learning algorithms using TCGA, GSE14764, GSE26193, GSE26712, GSE63885 and GSE140082 datasets. Gaussian mixture and logistic regression were performed to identify the optimal models for classifying lymphatic and venous invasion.

RESULTS

The optimal prognostic PRS developed by Lasso + survivalSVM algorithm acted as an independent risk factor for overall survival (OS) of ovarian cancer patients and had a good performance in evaluating OS rate of ovarian cancer patients. Significant correlation was obtained between PRS-based risk score and Immune score, ESTIMATE score, immune cells and cancer-related hallmarks. Low risk score indicated a lower immune escape score, TIDE score, and higher PD1&CTLA4 immunophenoscore in ovarian cancer. Moreover, PRS-based risk score acted as an indicator for drug sensitivity in the immunotherapy and chemotherapy of ovarian cancer patients.

CONCLUSIONS

All in all, our study developed a prognostic PRS showing powerful and good performance in predicting clinical outcome of ovarian cancer patients. PRS could serve as an indicator for drug sensitivity in the chemotherapy and immunotherapy.

Vis-SPLIT: Interactive Hierarchical Modeling for mRNA Expression Classification

We propose an interactive visual analytics tool, Vis-SPLIT, for partitioning a population of individuals into groups with similar gene signatures. Vis-SPLIT allows users to interactively explore a dataset and exploit visual separations to build a classification model for specific cancers. The visualization components reveal gene expression and correlation to assist specific partitioning decisions, while also providing overviews for the decision model and clustered genetic signatures. We demonstrate the effectiveness of our framework through a case study and evaluate its usability with domain experts. Our results show that Vis-SPLIT can classify patients based on their genetic signatures to effectively gain insights into RNA sequencing data, as compared to an existing classification system.

Evaluation and Exploration of Machine Learning and Convolutional Neural Network Classifiers in Detection of Lung Cancer from Microarray Gene-A Paradigm Shift

Microarray gene expression-based detection and classification of medical conditions have been prominent in research studies over the past few decades. However, extracting relevant data from the high-volume microarray gene expression with inherent nonlinearity and inseparable noise components raises significant challenges during data classification and disease detection. The dataset used for the research is the Lung Harvard 2 Dataset (LH2) which consists of 150 Adenocarcinoma subjects and 31 Mesothelioma subjects. The paper proposes a two-level strategy involving feature extraction and selection methods before the classification step. The feature extraction step utilizes Short Term Fourier Transform (STFT), and the feature selection step employs Particle Swarm Optimization (PSO) and Harmonic Search (HS) metaheuristic methods. The classifiers employed are Nonlinear Regression, Gaussian Mixture Model, Softmax Discriminant, Naive Bayes, SVM (Linear), SVM (Polynomial), and SVM (RBF). The two-level extracted relevant features are compared with raw data classification results, including Convolutional Neural Network (CNN) methodology. Among the methods, STFT with PSO feature selection and SVM (RBF) classifier produced the highest accuracy of 94.47%.

Identification of biomarkers associated with diagnosis of postmenopausal osteoporosis patients based on bioinformatics and machine learning

Background: Accumulating evidence suggests that postmenopausal osteoporosis (PMOP) is a common chronic systemic metabolic bone disease, but its specific molecular pathogenesis remains unclear. This study aimed to identify novel genetic diagnostic markers for PMOP. Methods: In this paper, we combined three GEO datasets to identify differentially expressed genes (DEGs) and performed functional enrichment analysis of PMOP-related differential genes. Key genes were analyzed using two machine learning algorithms, namely, LASSO and the Gaussian mixture model, and candidate biomarkers were found after taking the intersection. After further ceRNA network construction, methylation analysis, and immune infiltration analysis, ACACB and WWP1 were finally selected as diagnostic markers. Twenty-four clinical samples were collected, and the expression levels of biomarkers in PMOP were detected by qPCR. Results: We identified 34 differential genes in PMOP. DEG enrichment was mainly related to amino acid synthesis, inflammatory response, and apoptosis. The ceRNA network construction found that XIST-hsa-miR-15a-5p/hsa-miR-15b-5p/hsa-miR-497-5p and hsa-miR-195-5p-WWP1/ACACB may be RNA regulatory pathways regulating PMOP disease progression. ACACB and WWP1 were identified as diagnostic genes for PMOP, and validated in datasets and clinical sample experiments. In addition, these two genes were also significantly associated with immune cells, such as T, B, and NK cells. Conclusion: Overall, we identified two vital diagnostic genes responsible for PMOP. The results may help provide potential immunotherapeutic targets for PMOP.

Inferring cell diversity in single cell data using consortium-scale epigenetic data as a biological anchor for cell identity

Methods for cell clustering and gene expression from single-cell RNA sequencing (scRNA-seq) data are essential for biological interpretation of cell processes. Here, we present TRIAGE-Cluster which uses genome-wide epigenetic data from diverse bio-samples to identify genes demarcating cell diversity in scRNA-seq data. By integrating patterns of repressive chromatin deposited across diverse cell types with weighted density estimation, TRIAGE-Cluster determines cell type clusters in a 2D UMAP space. We then present TRIAGE-ParseR, a machine learning method which evaluates gene expression rank lists to define gene groups governing the identity and function of cell types. We demonstrate the utility of this two-step approach using atlases of in vivo and in vitro cell diversification and organogenesis. We also provide a web accessible dashboard for analysis and download of data and software. Collectively, genome-wide epigenetic repression provides a versatile strategy to define cell diversity and study gene regulation of scRNA-seq data.

RCFGL: Rapid Condition adaptive Fused Graphical Lasso and application to modeling brain region co-expression networks

Inferring gene co-expression networks is a useful process for understanding gene regulation and pathway activity. The networks are usually undirected graphs where genes are represented as nodes and an edge represents a significant co-expression relationship. When expression data of multiple (p) genes in multiple (K) conditions (e.g., treatments, tissues, strains) are available, joint estimation of networks harnessing shared information across them can significantly increase the power of analysis. In addition, examining condition-specific patterns of co-expression can provide insights into the underlying cellular processes activated in a particular condition. Condition adaptive fused graphical lasso (CFGL) is an existing method that incorporates condition specificity in a fused graphical lasso (FGL) model for estimating multiple co-expression networks. However, with computational complexity of O(p2K log K), the current implementation of CFGL is prohibitively slow even for a moderate number of genes and can only be used for a maximum of three conditions. In this paper, we propose a faster alternative of CFGL named rapid condition adaptive fused graphical lasso (RCFGL). In RCFGL, we incorporate the condition specificity into another popular model for joint network estimation, known as fused multiple graphical lasso (FMGL). We use a more efficient algorithm in the iterative steps compared to CFGL, enabling faster computation with complexity of O(p2K) and making it easily generalizable for more than three conditions. We also present a novel screening rule to determine if the full network estimation problem can be broken down into estimation of smaller disjoint sub-networks, thereby reducing the complexity further. We demonstrate the computational advantage and superior performance of our method compared to two non-condition adaptive methods, FGL and FMGL, and one condition adaptive method, CFGL in both simulation study and real data analysis. We used RCFGL to jointly estimate the gene co-expression networks in different brain regions (conditions) using a cohort of heterogeneous stock rats. We also provide an accommodating C and Python based package that implements RCFGL.

Semiparametric modelling of two-component mixtures with stochastic dominance

In this work, we studied a two-component mixture model with stochastic dominance constraint, a model arising naturally from many genetic studies. To model the stochastic dominance, we proposed a semiparametric modelling of the log of density ratio. More specifically, when the log of the ratio of two component densities is in a linear regression form, the stochastic dominance is immediately satisfied. For the resulting semiparametric mixture model, we proposed two estimators, maximum empirical likelihood estimator (MELE) and minimum Hellinger distance estimator (MHDE), and investigated their asymptotic properties such as consistency and normality. In addition, to test the validity of the proposed semiparametric model, we developed Kolmogorov-Smirnov type tests based on the two estimators. The finite-sample performance, in terms of both efficiency and robustness, of the two estimators and the tests were examined and compared via both thorough Monte Carlo simulation studies and real data analysis.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10463-022-00835-5.

EdgeCrafting: mining embedded, latent, nonlinear patterns to construct gene relationship networks

The mechanisms that coordinate cellular gene expression are highly complex and intricately interconnected. Thus, it is necessary to move beyond a fully reductionist approach to understanding genetic information flow and begin focusing on the networked connections between genes that organize cellular function. Continued advancements in computational hardware, coupled with the development of gene correlation network algorithms, provide the capacity to study networked interactions between genes rather than their isolated functions. For example, gene coexpression networks are used to construct gene relationship networks using linear metrics such as Spearman or Pearson correlation. Recently, there have been tools designed to deepen these analyses by differentiating between intrinsic vs extrinsic noise within gene expression values, identifying different modules based on tissue phenotype, and capturing potential nonlinear relationships. In this report, we introduce an algorithm with a novel application of image-based segmentation modalities utilizing blob detection techniques applied for detecting bigenic edges in a gene expression matrix. We applied this algorithm called EdgeCrafting to a bulk RNA-sequencing gene expression matrix comprised of a healthy kidney and cancerous kidney data. We then compared EdgeCrafting against 4 other RNA expression analysis techniques: Weighted Gene Correlation Network Analysis, Knowledge Independent Network Construction, NetExtractor, and Differential gene expression analysis.

Identification of condition-specific biomarker systems in uterine cancer

Uterine cancer is the fourth most common cancer among women, projected to affect 66,000 US women in 2021. Uterine cancer often arises in the inner lining of the uterus, known as the endometrium, but can present as several different types of cancer, including endometrioid cancer, serous adenocarcinoma, and uterine carcinosarcoma. Previous studies have analyzed the genetic changes between normal and cancerous uterine tissue to identify specific genes of interest, including TP53 and PTEN. Here we used Gaussian Mixture Models to build condition-specific gene coexpression networks for endometrial cancer, uterine carcinosarcoma, and normal uterine tissue. We then incorporated uterine regulatory edges and investigated potential coregulation relationships. These networks were further validated using differential expression analysis, functional enrichment, and a statistical analysis comparing the expression of transcription factors and their target genes across cancerous and normal uterine samples. These networks allow for a more comprehensive look into the biological networks and pathways affected in uterine cancer compared with previous singular gene analyses. We hope this study can be incorporated into existing knowledge surrounding the genetics of uterine cancer and soon become clinical biomarkers as a tool for better prognosis and treatment.

Addressing noise in co-expression network construction

Gene co-expression networks (GCNs) provide multiple benefits to molecular research including hypothesis generation and biomarker discovery. Transcriptome profiles serve as input for GCN construction and are derived from increasingly larger studies with samples across multiple experimental conditions, treatments, time points, genotypes, etc. Such experiments with larger numbers of variables confound discovery of true network edges, exclude edges and inhibit discovery of context (or condition) specific network edges. To demonstrate this problem, a 475-sample dataset is used to show that up to 97% of GCN edges can be misleading because correlations are false or incorrect. False and incorrect correlations can occur when tests are applied without ensuring assumptions are met, and pairwise gene expression may not meet test assumptions if the expression of at least one gene in the pairwise comparison is a function of multiple confounding variables. The ‘one-size-fits-all’ approach to GCN construction is therefore problematic for large, multivariable datasets. Recently, the Knowledge Independent Network Construction toolkit has been used in multiple studies to provide a dynamic approach to GCN construction that ensures statistical tests meet assumptions and confounding variables are addressed. Additionally, it can associate experimental context for each edge of the network resulting in context-specific GCNs (csGCNs). To help researchers recognize such challenges in GCN construction, and the creation of csGCNs, we provide a review of the workflow.

Screening of key biomarkers of tendinopathy based on bioinformatics and machine learning algorithms

Tendinopathy is a complex multifaceted tendinopathy often associated with overuse and with its high prevalence resulting in significant health care costs. At present, the pathogenesis and effective treatment of tendinopathy are still not sufficiently elucidated. The purpose of this research is to intensely explore the genes, functional pathways, and immune infiltration characteristics of the occurrence and development of tendinopathy. The gene expression profile of GSE106292, GSE26051 and GSE167226 are downloaded from GEO (NCBI comprehensive gene expression database) and analyzed by WGCNA software bag using R software, GSE26051, GSE167226 data set is combined to screen the differential gene analysis. We subsequently performed gene enrichment analysis of Gene Ontology (GO) and "Kyoto Encyclopedia of Genes and Genomes" (KEGG), and immune cell infiltration analysis. By constructing the LASSO regression model, Support vector machine (SVM-REF) and Gaussian mixture model (GMMs) algorithms are used to screen, to identify early diagnostic genes. We have obtained a total of 171 DEGs through WGCNA analysis and differentially expressed genes (DEGs) screening. By GO and KEGG enrichment analysis, it is found that these dysregulated genes were related to mTOR, HIF-1, MAPK, NF-κB and VEGF signaling pathways. Immune infiltration analysis showed that M1 macrophages, activated mast cells and activated NK cells had infiltration significance. After analysis of THE LASSO SVM-REF and GMMs algorithms, we found that the gene MACROD1 may be a gene for early diagnosis. We identified the potential of tendon disease early diagnosis way and immune gene regulation MACROD1 key infiltration characteristics based on comprehensive bioinformatics analysis. These hub genes and functional pathways may as early biomarkers of tendon injuries and molecular therapy level target is used to guide drug and basic research.

An integrated approach to identify bimodal genes associated with prognosis in câncer

Bimodal gene expression (where a gene expression distribution has two maxima) is associated with phenotypic diversity in different biological systems. A critical issue, thus, is the integration of expression and phenotype data to identify genuine associations. Here, we developed tools that allow both: i) the identification of genes with bimodal gene expression and ii) their association with prognosis in cancer patients from The Cancer Genome Atlas (TCGA). Bimodality was observed for 554 genes in expression data from 25 tumor types. Furthermore, 96 of these genes presented different prognosis when patients belonging to the two expression peaks were compared. The software to execute the method and the corresponding documentation are available at the Data access section.

Named Data Networking for Genomics Data Management and Integrated Workflows

Advanced imaging and DNA sequencing technologies now enable the diverse biology community to routinely generate and analyze terabytes of high resolution biological data. The community is rapidly heading toward the petascale in single investigator laboratory settings. As evidence, the single NCBI SRA central DNA sequence repository contains over 45 petabytes of biological data. Given the geometric growth of this and other genomics repositories, an exabyte of mineable biological data is imminent. The challenges of effectively utilizing these datasets are enormous as they are not only large in the size but also stored in geographically distributed repositories in various repositories such as National Center for Biotechnology Information (NCBI), DNA Data Bank of Japan (DDBJ), European Bioinformatics Institute (EBI), and NASA’s GeneLab. In this work, we first systematically point out the data-management challenges of the genomics community. We then introduce Named Data Networking (NDN), a novel but well-researched Internet architecture, is capable of solving these challenges at the network layer. NDN performs all operations such as forwarding requests to data sources, content discovery, access, and retrieval using content names (that are similar to traditional filenames or filepaths) and eliminates the need for a location layer (the IP address) for data management. Utilizing NDN for genomics workflows simplifies data discovery, speeds up data retrieval using in-network caching of popular datasets, and allows the community to create infrastructure that supports operations such as creating federation of content repositories, retrieval from multiple sources, remote data subsetting, and others. Named based operations also streamlines deployment and integration of workflows with various cloud platforms. Our contributions in this work are as follows 1) we enumerate the cyberinfrastructure challenges of the genomics community that NDN can alleviate, and 2) we describe our efforts in applying NDN for a contemporary genomics workflow (GEMmaker) and quantify the improvements. The preliminary evaluation shows a sixfold speed up in data insertion into the workflow. 3) As a pilot, we have used an NDN naming scheme (agreed upon by the community and discussed in Section 4) to publish data from broadly used data repositories including the NCBI SRA. We have loaded the NDN testbed with these pre-processed genomes that can be accessed over NDN and used by anyone interested in those datasets. Finally, we discuss our continued effort in integrating NDN with cloud computing platforms, such as the Pacific Research Platform (PRP). The reader should note that the goal of this paper is to introduce NDN to the genomics community and discuss NDN’s properties that can benefit the genomics community. We do not present an extensive performance evaluation of NDN—we are working on extending and evaluating our pilot deployment and will present systematic results in a future work.

The Role of Network Science in Glioblastoma

Network science has long been recognized as a well-established discipline across many biological domains. In the particular case of cancer genomics, network discovery is challenged by the multitude of available high-dimensional heterogeneous views of data. Glioblastoma (GBM) is an example of such a complex and heterogeneous disease that can be tackled by network science. Identifying the architecture of molecular GBM networks is essential to understanding the information flow and better informing drug development and pre-clinical studies. Here, we review network-based strategies that have been used in the study of GBM, along with the available software implementations for reproducibility and further testing on newly coming datasets. Promising results have been obtained from both bulk and single-cell GBM data, placing network discovery at the forefront of developing a molecularly-informed-based personalized medicine.

The role of gene to gene interaction in the breast's genomic signature of pregnancy

Full-term pregnancy at an early age confers long-term protection against breast cancer. Published data shows a specific transcriptomic profile controlling chromatin remodeling that could play a relevant role in the pregnancy-induced protection. This process of chromatin remodeling, induced by the breast differentiation caused by the first full-term pregnancy, has mainly been measured by the expression level of genes individually considered. However, genes equally expressed during the process of chromatin remodeling may behave differently in their interaction with other genes. These changes at the gene cluster level could constitute an additional dimension of chromatin remodeling and therefore of the pregnancy-induced protection. In this research, we apply Information and Graph Theories, Differential Co-expression Network Analysis, and Multiple Regression Analysis, specially designed to examine structural and informational aspects of data sets, to analyze this question. Our findings demonstrate that, independently of the changes in the gene expression at the individual level, there are significant changes in gene-gene interactions and gene cluster behaviors. These changes indicate that the parous breast, through the process of early full-term pregnancy, generates more modules in the networks, with higher density, and a genomic structure performing additional and more complex functions than those found in the nulliparous breast.

Pan-cancer driver copy number alterations identified by joint expression/CNA data analysis

Analysis of large gene expression datasets from biopsies of cancer patients can identify co-expression signatures representing particular biomolecular events in cancer. Some of these signatures involve genomically co-localized genes resulting from the presence of copy number alterations (CNAs), for which analysis of the expression of the underlying genes provides valuable information about their combined role as oncogenes or tumor suppressor genes. Here we focus on the discovery and interpretation of such signatures that are present in multiple cancer types due to driver amplifications and deletions in particular regions of the genome after doing a comprehensive analysis combining both gene expression and CNA data from The Cancer Genome Atlas.

Exploration into biomarker potential of region-specific brain gene co-expression networks

The human brain is a complex organ that consists of several regions each with a unique gene expression pattern. Our intent in this study was to construct a gene co-expression network (GCN) for the normal brain using RNA expression profiles from the Genotype-Tissue Expression (GTEx) project. The brain GCN contains gene correlation relationships that are broadly present in the brain or specific to thirteen brain regions, which we later combined into six overarching brain mini-GCNs based on the brain's structure. Using the expression profiles of brain region-specific GCN edges, we determined how well the brain region samples could be discriminated from each other, visually with t-SNE plots or quantitatively with the Gene Oracle deep learning classifier. Next, we tested these gene sets on their relevance to human tumors of brain and non-brain origin. Interestingly, we found that genes in the six brain mini-GCNs showed markedly higher mutation rates in tumors relative to matched sets of random genes. Further, we found that cortex genes subdivided Head and Neck Squamous Cell Carcinoma (HNSC) tumors and Pheochromocytoma and Paraganglioma (PCPG) tumors into distinct groups. The brain GCN and mini-GCNs are useful resources for the classification of brain regions and identification of biomarker genes for brain related phenotypes.

NetExtractor: Extracting a Cerebellar Tissue Gene Regulatory Network Using Differentially Expressed High Mutual Information Binary RNA Profiles

Bigenic expression relationships are conventionally defined based on metrics such as Pearson or Spearman correlation that cannot typically detect latent, non-linear dependencies or require the relationship to be monotonic. Further, the combination of intrinsic and extrinsic noise as well as embedded relationships between sample sub-populations reduces the probability of extracting biologically relevant edges during the construction of gene co-expression networks (GCNs). In this report, we address these problems via our NetExtractor algorithm. NetExtractor examines all pairwise gene expression profiles first with Gaussian mixture models (GMMs) to identify sample sub-populations followed by mutual information (MI) analysis that is capable of detecting non-linear differential bigenic expression relationships. We applied NetExtractor to brain tissue RNA profiles from the Genotype-Tissue Expression (GTEx) project to obtain a brain tissue specific gene expression relationship network centered on cerebellar and cerebellar hemisphere enriched edges. We leveraged the PsychENCODE pre-frontal cortex (PFC) gene regulatory network (GRN) to construct a cerebellar cortex (cerebellar) GRN associated with transcriptionally active regions in cerebellar tissue. Thus, we demonstrate the utility of our NetExtractor approach to detect biologically relevant and novel non-linear binary gene relationships.

A panel of eight microRNAs is a good predictive parameter for triple-negative breast cancer relapse

Rationale: Triple-negative breast cancer (TNBC), which has the highest recurrence rate and shortest survival time of all breast cancers, is in urgent need of a risk assessment method to determine an accurate treatment course. Recently, miRNA expression patterns have been identified as potential biomarkers for diagnosis, prognosis, and personalized therapy. Here, we investigate a combination of candidate miRNAs as a clinically applicable signature that can precisely predict relapse in TNBC patients after surgery. Methods: Four total cohorts of training (TCGA_TNBC and GEOD-40525) and validation (GSE40049 and GSE19783) datasets were analyzed with logistic regression and Gaussian mixture analyses. We established a miRNA signature risk model and identified an 8-miRNA signature for the prediction of TNBC relapse. Results: The miRNA signature risk model identified ten candidate miRNAs in the training set. By combining 8 of the 10 miRNAs (miR-139-5p, miR-10b-5p, miR-486-5p, miR-455-3p, miR-107, miR-146b-5p, miR-324-5p and miR-20a-5p), an accurate predictive model of relapse in TNBC patients was established and was highly correlated with prognosis (AUC of 0.80). Subsequently, this 8-miRNA signature prognosticated relapse in the two validation sets with AUCs of 0.89 and 0.90. Conclusion: The 8-miRNA signature predictive model may help clinicians provide a prognosis for TNBC patients with a high risk of recurrence after surgery and provide further personalized treatment to decrease the chance of relapse.

Joint Transcriptomic Analysis of Lung Cancer and Other Lung Diseases

Background: Epidemiological and clinical evidence points cancer comorbidity with pulmonary chronic disease. The acquisition of some hallmarks of cancer by cells affected with lung pathologies as a cell adaptive mechanism to a shear stress, suggests that could be associated with the establishment of tumoral processes. Objective: To propose a bioinformatic pipeline for the identification of all deregulated genes and the transcriptional regulators (TFs) that are coexpressed during lung cancer establishment, and therefore could be important for the acquisition of the hallmarks of cancer. Methods: Ten microarray datasets (six of lung cancer, four of lung diseases) comparing normal and diseases-related lung tissue were selected to identify hub differentiated expressed genes (DEGs) in common between lung pathologies and lung cancer, along with transcriptional regulators through the utilization of specialized libraries from R language. DAVID bioinformatics tool for gene enrichment analyses was used to identify genes with experimental evidence associated to tumoral processes and signaling pathways. Coexpression networks of DEGs and TFs in lung cancer establishment were created with Coexnet library, and a survival analysis of the main hub genes was made. Results: Two hundred ten DEGs were identified in common between lung cancer and other lung diseases related to the acquisition of tumoral characteristics, which are coexpressed in a lung cancer network with TFs, suggesting that could be related to the establishment of the tumoral pathology in lung. The comparison of the coexpression networks of lung cancer and other lung diseases allowed the identification of common connectivity patterns (CCPs) with DEGs and TFs correlated to important tumoral processes and signaling pathways, that haven´t been studied to experimentally validate their role in the early stages of lung cancer. Some of the TFs identified showed a correlation between its expression levels and the survival of lung cancer patients. Conclusion: Our findings indicate that lung diseases share genes with lung cancer which are coexpressed in lung cancer, and might be able to explain the epidemiological observations that point to direct and inverse comorbid associations between some chronic lung diseases and lung cancer and represent a complex transcriptomic scenario.

Identifying Temporally Regulated Root Nodulation Biomarkers Using Time Series Gene Co-Expression Network Analysis

Root nodulation results from a symbiotic relationship between a plant host and Rhizobium bacteria. Synchronized gene expression patterns over the course of rhizobial infection result in activation of pathways that are unique but overlapping with the highly conserved pathways that enable mycorrhizal symbiosis. We performed RNA sequencing of 30 Medicago truncatula root maturation zone samples at five distinct time points. These samples included plants inoculated with Sinorhizobium medicae and control plants that did not receive any Rhizobium. Following gene expression quantification, we identified 1,758 differentially expressed genes at various time points. We constructed a gene co-expression network (GCN) from the same data and identified link community modules (LCMs) that were comprised entirely of differentially expressed genes at specific time points post-inoculation. One LCM included genes that were up-regulated at 24 h following inoculation, suggesting an activation of allergen family genes and carbohydrate-binding gene products in response to Rhizobium. We also identified two LCMs that were comprised entirely of genes that were down regulated at 24 and 48 h post-inoculation. The identity of the genes in these modules suggest that down-regulating specific genes at 24 h may result in decreased jasmonic acid production with an increase in cytokinin production. At 48 h, coordinated down-regulation of a specific set of genes involved in lipid biosynthesis may play a role in nodulation. We show that GCN-LCM analysis is an effective method to preliminarily identify polygenic candidate biomarkers of root nodulation and develop hypotheses for future discovery.

EdgeScaping: Mapping the spatial distribution of pairwise gene expression intensities

Gene co-expression networks (GCNs) are constructed from Gene Expression Matrices (GEMs) in a bottom up approach where all gene pairs are tested for correlation within the context of the input sample set. This approach is computationally intensive for many current GEMs and may not be scalable to millions of samples. Further, traditional GCNs do not detect non-linear relationships missed by correlation tests and do not place genetic relationships in a gene expression intensity context. In this report, we propose EdgeScaping, which constructs and analyzes the pairwise gene intensity network in a holistic, top down approach where no edges are filtered. EdgeScaping uses a novel technique to convert traditional pairwise gene expression data to an image based format. This conversion not only performs feature compression, making our algorithm highly scalable, but it also allows for exploring non-linear relationships between genes by leveraging deep learning image analysis algorithms. Using the learned embedded feature space we implement a fast, efficient algorithm to cluster the entire space of gene expression relationships while retaining gene expression intensity. Since EdgeScaping does not eliminate conventionally noisy edges, it extends the identification of co-expression relationships beyond classically correlated edges to facilitate the discovery of novel or unusual expression patterns within the network. We applied EdgeScaping to a human tumor GEM to identify sets of genes that exhibit conventional and non-conventional interdependent non-linear behavior associated with brain specific tumor sub-types that would be eliminated in conventional bottom-up construction of GCNs. Edgescaping source code is available at https://github.com/bhusain/EdgeScaping under the MIT license.

Moving Just Enough Deep Sequencing Data to Get the Job Done

Motivation:

As the size of high-throughput DNA sequence datasets continues to grow, the cost of transferring and storing the datasets may prevent their processing in all but the largest data centers or commercial cloud providers. To lower this cost, it should be possible to process only a subset of the original data while still preserving the biological information of interest.

Results:

Using 4 high-throughput DNA sequence datasets of differing sequencing depth from 2 species as use cases, we demonstrate the effect of processing partial datasets on the number of detected RNA transcripts using an RNA-Seq workflow. We used transcript detection to decide on a cutoff point. We then physically transferred the minimal partial dataset and compared with the transfer of the full dataset, which showed a reduction of approximately 25% in the total transfer time. These results suggest that as sequencing datasets get larger, one way to speed up analysis is to simply transfer the minimal amount of data that still sufficiently detects biological signal.

Availability:

All results were generated using public datasets from NCBI and publicly available open source software.

Gene co-expression network analysis identifies trait-related modules in Arabidopsis thaliana

A comprehensive network of the Arabidopsis transcriptome was analyzed and may serve as a valuable resource for candidate gene function investigations. A web tool to explore module information was also provided. Arabidopsis thaliana is a widely studied model plant whose transcriptome has been substantially profiled in various tissues, development stages and other conditions. These data can be reused for research on gene function through a systematic analysis of gene co-expression relationships. We collected microarray data from National Center for Biotechnology Information Gene Expression Omnibus, identified modules of co-expressed genes and annotated module functions. These modules were associated with experiments/traits, which provided potential signature modules for phenotypes. Novel heat shock proteins were implicated according to guilt by association. A higher-order module networks analysis suggested that the Arabidopsis network can be further organized into 15 meta-modules and that a chloroplast meta-module has a distinct gene expression pattern from the other 14 meta-modules. A comparison with the rice transcriptome revealed preserved modules and KEGG pathways. All the module gene information was available from an online tool at http://bioinformatics.fafu.edu.cn/arabi/ . Our findings provide a new source for future gene discovery in Arabidopsis.

Linking Binary Gene Relationships to Drivers of Renal Cell Carcinoma Reveals Convergent Function in Alternate Tumor Progression Paths

Renal cell carcinoma (RCC) subtypes are characterized by distinct molecular profiles. Using RNA expression profiles from 1,009 RCC samples, we constructed a condition-annotated gene coexpression network (GCN). The RCC GCN contains binary gene coexpression relationships (edges) specific to conditions including RCC subtype and tumor stage. As an application of this resource, we discovered RCC GCN edges and modules that were associated with genetic lesions in known RCC driver genes, including VHL, a common initiating clear cell RCC (ccRCC) genetic lesion, and PBRM1 and BAP1 which are early genetic lesions in the Braided Cancer River Model (BCRM). Since ccRCC tumors with PBRM1 mutations respond to targeted therapy differently than tumors with BAP1 mutations, we focused on ccRCC-specific edges associated with tumors that exhibit alternate mutation profiles: VHL-PBRM1 or VHL-BAP1. We found specific blends molecular functions associated with these two mutation paths. Despite these mutation-associated edges having unique genes, they were enriched for the same immunological functions suggesting a convergent functional role for alternate gene sets consistent with the BCRM. The condition annotated RCC GCN described herein is a novel data mining resource for the assignment of polygenic biomarkers and their relationships to RCC tumors with specific molecular and mutational profiles.

Sorting Five Human Tumor Types Reveals Specific Biomarkers and Background Classification Genes

We applied two state-of-the-art, knowledge independent data-mining methods - Dynamic Quantum Clustering (DQC) and t-Distributed Stochastic Neighbor Embedding (t-SNE) - to data from The Cancer Genome Atlas (TCGA). We showed that the RNA expression patterns for a mixture of 2,016 samples from five tumor types can sort the tumors into groups enriched for relevant annotations including tumor type, gender, tumor stage, and ethnicity. DQC feature selection analysis discovered 48 core biomarker transcripts that clustered tumors by tumor type. When these transcripts were removed, the geometry of tumor relationships changed, but it was still possible to classify the tumors using the RNA expression profiles of the remaining transcripts. We continued to remove the top biomarkers for several iterations and performed cluster analysis. Even though the most informative transcripts were removed from the cluster analysis, the sorting ability of remaining transcripts remained strong after each iteration. Further, in some iterations we detected a repeating pattern of biological function that wasn't detectable with the core biomarker transcripts present. This suggests the existence of a "background classification" potential in which the pattern of gene expression after continued removal of "biomarker" transcripts could still classify tumors in agreement with the tumor type.

Discovery and validation of a glioblastoma co-expressed gene module

Tumors exhibit complex patterns of aberrant gene expression. Using a knowledge-independent, noise-reducing gene co-expression network construction software called KINC, we created multiple RNAseq-based gene co-expression networks relevant to brain and glioblastoma biology. In this report, we describe the discovery and validation of a glioblastoma-specific gene module that contains 22 co-expressed genes. The genes are upregulated in glioblastoma relative to normal brain and lower grade glioma samples; they are also hypo-methylated in glioblastoma relative to lower grade glioma tumors. Among the proneural, neural, mesenchymal, and classical glioblastoma subtypes, these genes are most-highly expressed in the mesenchymal subtype. Furthermore, high expression of these genes is associated with decreased survival across each glioblastoma subtype. These genes are of interest to glioblastoma biology and our gene interaction discovery and validation workflow can be used to discover and validate co-expressed gene modules derived from any co-expression network.