According to a recent disease statistic report released by American Heart Association (AHA), cardiovascular diseases (CVD) remain the leading cause of death in America (Benjamin et al., 2017). The treatment of CVD also imposed a huge economic burden on the healthcare system (Benjamin et al., 2017). Deep understanding of the mechanism of CVD is a valuable approach for devising effective novel cardiovascular therapeutics.

With increasing number of transcriptomic studies (including microarray and RNA-sequencing) performed in cultured cells as well as in experimental mice or patients with CVD, we now have the capability to understand the influence of therapeutic intervention or gene perturbation on CVD outcome at genome-wide levels which were inaccessible in the past. However, the value of these transcriptomic data was always underestimated since most of the deposited data are not released to public until manuscripts are published. Therefore, it is critical to make large-scale efforts to mine, validate, and integrate the underlying information streams arising from various transcriptomics studies (Musunuru et al., 2017). To meet the increasing need of precision medicine, AHA has recently established the Institute for Precision Cardiovascular Medicine¹, offering a new category of data-mining grants focused on harmonizing and mining CVD-based data for cardiovascular therapeutics. Therefore, in this article, I will summarize the workflow of transcriptomic profiling, basic bioinformatics analysis, and those profiling studies performed in vascular cells as well as human and mice diseased samples, aiming to provide a direct resource gallery in systems vascular medicine. Obviously, further mining of these publicly available datasets will provide a useful resource for understanding the cellular basis of atherosclerotic vascular diseases.

For analyzing a small number of gene transcripts, quantitative real-time PCR or pathway-focused (such as pathways of angiogenesis or endothelial cell biology) gene expression analysis using PCR arrays (such as RT² Profiler PCR Arrays from Qiagen) can be used. In order to understand genome-wide influence of different conditions on CVD outcome, DNA microarray and RNA-sequencing (RNA-seq) are frequently used. Traditional transcriptomic analysis was mostly performed by using DNA microarray, which employs dye (Cy3, Cy5) hybridization-based technology to analyze differential gene expression pattern under certain conditions (such as gene knockout, or drug/stimuli treatment), although microarray has several technical limitations (de Franciscis et al., 2016; Haase et al., 2016). Recently, with the advent of next-generation sequencing technology, transcriptomic analysis has transitioned to RNA-seq (Wang et al., 2009), to quantify the amount of transcripts including protein-coding genes (mRNA), splice variants, as well as long non-coding RNA transcripts (lncRNA) in biological samples at genome-wide level (Mortazavi et al., 2008). Comparatively speaking, RNA-seq has the capability to identify more differentially expressed genes in various cell types than gene microarray (Wang et al., 2009; Zhang et al., 2014). In addition, there are also some commercial lncRNA array services available, such as Arraystar LncRNA Expression Arrays² which systematically profile lncRNAs together with protein-coding mRNAs. A typical workflow of transcriptomic analysis involves several steps: (1) sample preparation; (2) RNA isolation by TRIzol or other commercial kits; (3) high-quality RNA submitted to Core facility or commercial companies for RNA-seq; or reverse transcription to cDNA for hybridization-based microarray analysis (Figure 1). To visualize the result of data analysis, gene expression values from both transcriptomic analyses can be represented as heat maps, listing the most significantly changed genes in assays. Downstream analysis of microarray and RNA-seq are quite similar, include gene ontology (GO) enrichment and pathway analysis as well as functionally classification of gene annotation (Yue and Reisdorf, 2005).

Currently, microarrays remain a widely used approach for transcriptome studies due to its relatively low cost (readily affordable by many researchers) and ease to process large numbers of samples (Zhao et al., 2014). However, microarray has several limitations, most of which arise from probe and hybridization-related issues (probe performance and non-specific hybridization etc), such as high background level, difficult to detect very lowly expressed transcripts, and novel transcripts as well as splice variants (Draghici et al., 2006; Zhao et al., 2014). In contrast, RNA-seq has obvious advantages in these aspects (Russo et al., 2003; Wang et al., 2009; Zhao et al., 2014; Zhang et al., 2015) (Table 1): (1) Ability to detect novel transcripts; (2) Wider dynamic range of detection; (3) High signal-to-noise ratio; (4) High reproducibility and low variation. However, performing RNA-seq-based experiments is more expensive than microarray-based experiments, and requires extensive technical and bioinformatic expertise in data analysis (Zhao et al., 2014). The cost issue would potentially limit its application in experimental and clinical medicine. Moreover, a typical RNA-seq data in various formats was at the scale of GB depending on the number of samples tested. This presents a potential challenge for RAW data storage, processing, and analysis (Draghici et al., 2006). Fortunately, with recent technological advances, the costs for performing sequencing have declined; thus, RNA-seq is becoming more affordable than usual to users. Also various data depositing platforms (such as Gene Expression Ominbus and ArrayExpress) have emerged, and these platforms significantly solved the storage issue of large-scale RNA-seq RAW data. Readers are referred to references (Russo et al., 2003; Draghici et al., 2006; Wang et al., 2009; Zhao et al., 2014; Zhang et al., 2015) for details of the advantages and limitations of RNA-seq and microarray technology.

Traditionally, transcriptomic data were included as supplementary information in published scientific literature. Nowadays, to meet the need of open data and data sharing, most of the transcriptomic profiling data were deposited in ArrayExpress³ and NCBI Gene Expression Ominbus (GEO)⁴. In this study, I will summarize part of the datasets that has been deposited in GEO database with supported publication records.

Currently, there are many softwares or websites that can help researchers analyze the data obtained from microarray and RNA-seq when uploading gene ID list. I summarize here some of the softwares and websites in Table 2. Basic bioinformatics analysis of transcriptomic data include the following (Yue and Reisdorf, 2005):

Due to the fact that gene microarray or RNA-seq generates big data that cannot be presented in regular format, therefore, it is critical to mine the data/information deposited in publicly available databases and perform related analysis.

Transcriptomic comparisons would facilitate the identification of differentially expressed transcripts between human diseased and control samples, in different vascular cell types (endothelial cells, monocytes/macrophages, and smooth muscle cells), or in response to different pharmacological/genetic/environmental perturbations (Musunuru et al., 2017). Three common types of transcriptome profiling in vascular biology are summarized as below (Tables 3–5):

Genome-wide transcriptomic analysis by microarray and RNA-seq emerge as powerful tools for translational research. Serval studies have conducted microarray and RNA-seq in parallel to compare the similarity and difference in transcriptome profiling of target cells/tissues and found that there is a high concordance between two technologies, suggesting the general suitability and reliability of both technologies (Bottomly et al., 2011; Sirbu et al., 2012; Zhao et al., 2014). Both technologies greatly transformed our traditional method of research into “discovery”-based method for mechanistic studies, allowing us to readily evaluate the cell-type and stimulus (or small-molecule drug)-specific regulation of gene expression. From these studies, readers can also mine data according to web-based instructions, and develop a working hypothesis to test whether gene X is involved in the development and progression of vascular diseases. Due to the complex nature of transcriptional regulation, the level of specific transcripts is determined by both transcriptional machinery and environment (such as diet, nutrient etc) (Musunuru et al., 2017). Together with recently emergence of single cell RNA-seq (Linnarsson and Teichmann, 2016), these biotechnological advances will provide powerful toolboxes for understanding the vascular transcriptome and represent an import step toward precision cardiovascular medicine.

Overall, RNA-seq is a high-throughput, and “discovery-based” approach for investigating transcriptome of different samples. It is more sensitive in detecting very lowly expressed genes or extremely highly expressed genes, therefore, offers a wider dynamic range than traditional microarray (Zhao et al., 2014). It is replacing microarray and thus becoming the predominant tool for transcriptome studies in basic, translational and clinical research (Musunuru et al., 2017). Noteworthy, RNA-seq is also a powerful tool for investigating non-coding RNA, lncRNAs in particular [such as SMILR (Ballantyne et al., 2016), MANTIS (Leisegang et al., 2017), LincRNA-p21 (Wu et al., 2014), Lnc-Ang362 (Leung et al., 2013), MYOSLID (Zhao et al., 2016), RNCR3 (Shan et al., 2016)], regulates many facets of vascular biology (reviewed in Li et al., 2016; Poller et al., 2017), demystifying the “dark” genome of vascular diseases. These lncRNA emerge as important players and therapeutic targets in vascular diseases. However, we have to bear in mind that: different variables of experimental conditions (such as cell types, treatment time, and animal models) pose a challenge to make quantifiable conclusions from published datasets. At least, by mining and comparing different datasets from different studies, we can gain a general comprehension on specific genes that are commonly regulated by the same treatment/disease.

The invention of new datamining tools/softwares is a good strategy to mine existing data. Recently, a web-based tool, Transcriptomine (Becnel et al., 2017), was developed to mine data on dissecting the effects of genetic or pharmacological interventions on nuclear receptor signaling. We can envisage that deep mining of the “hiden” data in vascular medicine will definitely accelerate biomarker discovery and prompts the identification and functional characterization of novel therapeutic targets (including coding genes and lncRNAs) in vascular diseases including atherosclerosis, AAA, and other rare-occuring vascular diseases. To conclude, mining expression profiling data from bench to bedside represents a cost-efficient new avenue for research of precision cardiovascular medicine.

SX contributed to the conception, drafting, and revision of the manuscript and approved the final version.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.