Circular RNA (circRNA) is a novel class of single-stranded RNAs with a closed loop structure. The majority of circRNAs are formed by a back-splicing process in pre-mRNA splicing. Their expression is dynamically regulated and shows spatiotemporal patterns among cell types, tissues and developmental stages. CircRNAs have important biological functions in many physiological processes, and their aberrant expression is implicated in many human diseases. Due to their high stability, circRNAs are becoming promising biomarkers in many human diseases, such as cardiovascular diseases, autoimmune diseases and human cancers. In this review, we focus on the translational potential of using human blood circRNAs as liquid biopsy biomarkers for human diseases. We highlight their abundant expression, essential biological functions and significant correlations to human diseases in various components of peripheral blood, including whole blood, blood cells and extracellular vesicles. In addition, we summarize the current knowledge of blood circRNA biomarkers for disease diagnosis or prognosis.

Circular RNAs (circRNAs) are a class of novel RNAs with important biological functions, and aberrant expression of circRNAs has been implicated in human diseases. However, the feasibility of using blood circRNAs as disease biomarkers is largely unknown. We explored the potential of using human peripheral blood mononuclear cell (PBMC) circRNAs as marker molecules to diagnose active pulmonary tuberculosis (TB). First, we demonstrated that circRNAs are widely expressed in human PBMCs and that many are abundant enough to be detected. Second, we found that the magnitude of PBMC circRNAs in TB patients was higher than that in the paired healthy controls. Compared with host linear transcripts, the circRNAs within several pathways are disproportionately upregulated in active TB patients, including "Cytokine-cytokine receptor interaction", "Chemokine signaling pathway", "Neurotrophin signaling pathway", and "Bacterial invasion of epithelial cells". Based on the differentially expressed circRNAs within these pathways, we developed a PBMC circRNA-based molecular signature differentiating active TB patients from healthy controls. We validated the classification power of the PBMC circRNA signature in an independent cohort with the area under the receiver operating characteristic curve (AUC) at 0.946. Our results suggest that PBMC circRNAs are potentially reliable marker molecules in TB diagnosis.

One unmet challenge in lung cancer diagnosis is to accurately differentiate lung cancer from other lung diseases with similar clinical symptoms and radiological features, such as pulmonary tuberculosis (TB). To identify reliable biomarkers for lung cancer screening, we leverage the recently discovered non-canonical small non-coding RNAs (i.e., tRNA-derived small RNAs [tsRNAs], rRNA-derived small RNAs [rsRNAs], and YRNA-derived small RNAs [ysRNAs]) in human peripheral blood mononuclear cells and develop a molecular signature composed of distinct ts/rs/ysRNAs (TRY-RNA). Our TRY-RNA signature precisely discriminates between control, lung cancer, and pulmonary TB subjects in both the discovery and validation cohorts and outperforms microRNA-based biomarkers, which bears the diagnostic potential for lung cancer screening.

Circular RNAs (circRNAs) are a novel class of regulatory RNAs. Here, we present a comprehensive investigation of circRNA expression profiles across 11 tissues and four developmental stages in rats, along with cross-species analyses in humans and mice. Although the expression of circRNAs is positively correlated with that of cognate mRNAs, highly expressed genes tend to splice a larger fraction of circular transcripts. Moreover, circRNAs exhibit higher tissue specificity than cognate mRNAs. Intriguingly, while we observed a monotonic increase of circRNA abundance with age in the rat brain, we further discovered a dynamic, age-dependent pattern of circRNA expression in the testes that is characterized by a dramatic increase with advancing stages of sexual maturity and a decrease with aging. The age-sensitive testicular circRNAs are highly associated with spermatogenesis, independent of cognate mRNA expression. The tissue/age implications of circRNAs suggest that they present unique physiological functions rather than simply occurring as occasional by-products of gene transcription.

Hypoxia-inducible factor pathway genes are linked to adaptation in both human and nonhuman highland species. EPAS1 , a notable target of hypoxia adaptation, is associated with relatively lower hemoglobin concentration in Tibetans. We provide evidence for an association between an adaptive EPAS1 variant (rs570553380) and the same phenotype of relatively low hematocrit in Andean highlanders. This Andean-specific missense variant is present at a modest frequency in Andeans and absent in other human populations and vertebrate species except the coelacanth. CRISPR-base-edited human cells with this variant exhibit shifts in hypoxia-regulated gene expression, while metabolomic analyses reveal both genotype and phenotype associations and validation in a lowland population. Although this genocopy of relatively lower hematocrit in Andean highlanders parallels well-replicated findings in Tibetans, it likely involves distinct pathway responses based on a protein-coding versus noncoding variants, respectively. These findings illuminate how unique variants at EPAS1 contribute to the same phenotype in Tibetans and a subset of Andean highlanders despite distinct evolutionary trajectories.

Precise biomarker development is a key step in disease management. However, most of the published biomarkers were derived from a relatively small number of samples with supervised approaches. Recent advances in unsupervised machine learning promise to leverage very large datasets for making better predictions of disease biomarkers. Denoising autoencoder (DA) is one of the unsupervised deep learning algorithms, which is a stochastic version of autoencoder techniques. The principle of DA is to force the hidden layer of autoencoder to capture more robust features by reconstructing a clean input from a corrupted one. Here, a DA model was applied to analyze integrated transcriptomic data from 13 published lung cancer studies, which consisted of 1916 human lung tissue samples. Using DA, we discovered a molecular signature composed of multiple genes for lung adenocarcinoma (ADC). In independent validation cohorts, the proposed molecular signature is proved to be an effective classifier for lung cancer histological subtypes. Also, this signature successfully predicts clinical outcome in lung ADC, which is independent of traditional prognostic factors. More importantly, this signature exhibits a superior prognostic power compared with the other published prognostic genes. Our study suggests that unsupervised learning is helpful for biomarker development in the era of precision medicine.