CpG-Islands as Markers for Liquid Biopsies of Cancer Patients

Use of Multiple Machine Learning Approaches for Selecting Urothelial Cancer-Specific DNA Methylation Biomarkers in Urine

Diagnosing urothelial cancer (UCa) via invasive cystoscopy is painful, specifically in men, and can cause infection and bleeding. Because the UCa risk is higher for male patients, urinary non-invasive UCa biomarkers are highly desired to stratify men for invasive cystoscopy. We previously identified multiple DNA methylation sites in urine samples that detect UCa with a high sensitivity and specificity in men. Here, we identified the most relevant markers by employing multiple statistical approaches and machine learning (random forest, boosted trees, LASSO) using a dataset of 251 male UCa patients and 111 controls. Three CpG sites located in ALOX5, TRPS1 and an intergenic region on chromosome 16 have been concordantly selected by all approaches, and their combination in a single decision matrix for clinical use was tested based on their respective thresholds of the individual CpGs. The combination of ALOX5 and TRPS1 yielded the best overall sensitivity (61%) at a pre-set specificity of 95%. This combination exceeded both the diagnostic performance of the most sensitive bioinformatic approach and that of the best single CpG. In summary, we showed that overlap analysis of multiple statistical approaches identifies the most reliable biomarkers for UCa in a male collective. The results may assist in stratifying men for cystoscopy.

Bridging biological cfDNA features and machine learning approaches

Liquid biopsies (LBs), particularly using circulating tumor DNA (ctDNA), are expected to revolutionize precision oncology and blood-based cancer screening. Recent technological improvements, in combination with the ever-growing understanding of cell-free DNA (cfDNA) biology, are enabling the detection of tumor-specific changes with extremely high resolution and new analysis concepts beyond genetic alterations, including methylomics, fragmentomics, and nucleosomics. The interrogation of a large number of markers and the high complexity of data render traditional correlation methods insufficient. In this regard, machine learning (ML) algorithms are increasingly being used to decipher disease- and tissue-specific signals from cfDNA. Here, we review recent insights into biological ctDNA features and how these are incorporated into sophisticated ML applications.

Risk scoring based on DNA methylation-driven related DEGs for colorectal cancer prognosis with systematic insights

Colorectal cancer is a common malignant tumor of the digestive tract. Despite advances in diagnostic techniques and medications. Its prognosis remains challenging. DNA methylation-driven related circulating tumor cells have attracted enormous interest in diagnosing owing to their non-invasive nature and early recognition properties. However, the mechanism through which risk biomarkers act remains elusive. Here, we designed a risk model based on differentially expressed genes, DNA methylation, robust, and survival-related factors in the framework of Cox regression. The model has satisfactory performance and is independently verified by an external and isolated dataset in terms of C-index value, ROC, and tROC. The model was applied to Colorectal cancer patients who were subsequently divided into high- and low-risk groups. Functional annotations, genomic alterations, tumor immune environment, and drug sensitivity were analyzed. We observed that up-regulated genes are associated with epithelial cell differentiation and MAPK signaling pathways. The down-regulated genes are related to IL-7 signaling and apoptosis-induced DNA fragmentation. Interestingly, the immune system was inhibited in high-risk groups. High-frequency mutation genes tend to co-occur. High-risk score patients are related to copy number amplification events. To address the challenges, we suggested eleven and twenty-one drugs that are sensitive to low- and high-risk patients. Finally, an artificial neural network was provided to evaluate the immunotherapeutic efficiency. Taken together, the findings demonstrated that our risk score model is robust and reliable for evaluating the prognosis with novel diagnostic and treatment targets. It also yields benefits for the treatment and provides unique insights into developing therapeutic strategies.

Blood-based DNA methylation signatures in cancer: A systematic review

DNA methylation profiles are in dynamic equilibrium via the initiation of methylation, maintenance of methylation and demethylation, which control gene expression and chromosome stability. Changes in DNA methylation patterns play important roles in carcinogenesis and primarily manifests as hypomethylation of the entire genome and the hypermethylation of individual loci. These changes may be reflected in blood-based DNA, which provides a non-invasive means for cancer monitoring. Previous blood-based DNA detection objects primarily included circulating tumor DNA/cell-free DNA (ctDNA/cfDNA), circulating tumor cells (CTCs) and exosomes. Researchers gradually found that methylation changes in peripheral blood mononuclear cells (PBMCs) also reflected the presence of tumors. Blood-based DNA methylation is widely used in early diagnosis, prognosis prediction, dynamic monitoring after treatment and other fields of clinical research on cancer. The reversible methylation of genes also makes them important therapeutic targets. The present paper summarizes the changes in DNA methylation in cancer based on existing research and focuses on the characteristics of the detection objects of blood-based DNA, including ctDNA/cfDNA, CTCs, exosomes and PBMCs, and their application in clinical research.

Challenges in promoter methylation analysis in the new era of translational oncology: a focus on liquid biopsy

Toward the discovery of novel reliable biomarkers, epigenetic alterations have been repeatedly proposed for the diagnosis and the development of therapeutic strategies against cancer. Indeed, for promoter methylation to actively become a tumor marker for clinical use, it must be combined with a highly informative technology evaluated in an appropriate biospecimen. Methodological standardization related to epigenetic research is, in fact, one of the most challenging tasks. Moreover, tissue-based biopsy is being complemented and, in some cases, replaced by liquid biopsy. This review will highlight the advancements made for both pre-analytical and analytical implementation for the prospective use of methylation biomarkers in clinical settings, with particular emphasis on liquid biopsy.

Cancer Detection and Classification by CpG Island Hypermethylation Signatures in Plasma Cell-Free DNA

Simple Summary

The detection of DNA methylation changes in blood has emerged as a promising approach for cancer diagnosis and management. Our group has previously optimized a blood DNA methylation profiling technology that is based on affinity capture of methylated DNA, termed cfMBD-seq. The aim of this study was to assess the potential clinical feasibility of cfMBD-seq. We applied cfMBD-seq to the blood samples of cancer patients and identified methylation signatures that can not only discriminate cancer patients from cancer-free individuals but can also enable accurate multi-cancer classification. Our findings will help to expand on existing blood-based molecular diagnostic tests and identify novel methylation biomarkers for early cancer detection and classification.

Abstract

Cell-free DNA (cfDNA) methylation has emerged as a promising biomarker for early cancer detection, tumor type classification, and treatment response monitoring. Enrichment-based cfDNA methylation profiling methods such as cfMeDIP-seq have shown high accuracy in the classification of multiple cancer types. We have previously optimized another enrichment-based approach for ultra-low input cfDNA methylome profiling, termed cfMBD-seq. We reported that cfMBD-seq outperforms cfMeDIP-seq in the enrichment of high-CpG-density regions, such as CpG islands. However, the clinical feasibility of cfMBD-seq is unknown. In this study, we applied cfMBD-seq to profiling the cfDNA methylome using plasma samples from cancer patients and non-cancer controls. We identified 1759, 1783, and 1548 differentially hypermethylated CpG islands (DMCGIs) in lung, colorectal, and pancreatic cancer patients, respectively. Interestingly, the vast majority of DMCGIs were overlapped with aberrant methylation changes in corresponding tumor tissues, indicating that DMCGIs detected by cfMBD-seq were mainly driven by tumor-specific DNA methylation patterns. From the overlapping DMCGIs, we carried out machine learning analyses and identified a set of discriminating methylation signatures that had robust performance in cancer detection and classification. Overall, our study demonstrates that cfMBD-seq is a powerful tool for sensitive detection of tumor-derived epigenomic signals in cfDNA.

Impact of Preanalytical and Analytical Methods on Cell-Free DNA Diagnostics

While tissue biopsy has for the longest time been the gold-standard in biomedicine, precision/personalized medicine is making the shift toward liquid biopsies. Cell-free DNA (cfDNA) based genetic and epigenetic biomarkers reflect the molecular status of its tissue-of-origin allowing for early and non-invasive diagnostics of different pathologies. However, selection of preanalytical procedures (including cfDNA isolation) as well as analytical methods are known to impact the downstream results. Calls for greater standardization are made continuously, yet comprehensive assessments of the impact on diagnostic parameters are lacking. This study aims to evaluate the preanalytic and analytic factors that influence cfDNA diagnostic parameters in blood and semen. Text mining analysis has been performed to assess cfDNA research trends, and identify studies on isolation methods, preanalytical and analytical impact. Seminal and blood plasma were tested as liquid biopsy sources. Traditional methods of cfDNA isolation, commercial kits (CKs), and an in-house developed protocol were tested, as well as the impact of dithiothreitol (DTT) on cfDNA isolation performance. Fluorimetry, qPCR, digital droplet PCR (ddPCR), and bioanalyzer were compared as cfDNA quantification methods. Fragment analysis was performed by qPCR and bioanalyzer while the downstream application (cfDNA methylation) was analyzed by pyrosequencing. In contrast to blood, semen as a liquid biopsy source has only recently begun to be reported as a liquid biopsy source, with almost half of all publications on it being review articles. Experimental data revealed that cfDNA isolation protocols give a wide range of cfDNA yields, both from blood and seminal plasma. The addition of DTT to CKs has improved yields in seminal plasma and had a neutral/negative impact in blood plasma. Capillary electrophoresis and fluorometry reported much higher yields than PCR methods. While cfDNA yield and integrity were highly impacted, cfDNA methylation was not affected by isolation methodology or DTT. In conclusion, NucleoSnap was recognized as the kit with the best overall performance. DTT improved CK yields in seminal plasma. The in-house developed protocol has shown near-kit isolation performance. ddPCR LINE-1 assay for absolute detection of minute amounts of cfDNA was established and allowed for quantification of samples inhibited in qPCR. cfDNA methylation was recognized as a stable biomarker unimpacted by cfDNA isolation method. Finally, semen was found to be an abundant source of cfDNA offering potential research opportunities and benefits for cfDNA based biomarkers development related to male reproductive health.

Machine Learning Protocols in Early Cancer Detection Based on Liquid Biopsy: A Survey

With the advances of liquid biopsy technology, there is increasing evidence that body fluid such as blood, urine, and saliva could harbor the potential biomarkers associated with tumor origin. Traditional correlation analysis methods are no longer sufficient to capture the high-resolution complex relationships between biomarkers and cancer subtype heterogeneity. To address the challenge, researchers proposed machine learning techniques with liquid biopsy data to explore the essence of tumor origin together. In this survey, we review the machine learning protocols and provide corresponding code demos for the approaches mentioned. We discuss algorithmic principles and frameworks extensively developed to reveal cancer mechanisms and consider the future prospects in biomarker exploration and cancer diagnostics.

Cell-free DNA methylome profiling by MBD-seq with ultra-low input

Methylation signatures in cell-free DNA (cfDNA) have shown great sensitivity and specificity in the characterization of tumour status and classification of tumour types, as well as the response to therapy and recurrence. Currently, most cfDNA methylation studies are based on bisulphite conversion, especially targeted bisulphite sequencing, while enrichment-based methods such as cfMeDIP-seq are beginning to show potential. Here, we report an enrichment-based ultra-low input cfDNA methylation profiling method using methyl-CpG binding proteins capture, termed cfMBD-seq. We optimized the conditions for cfMBD capture by adjusting the amount of MethylCap protein along with using methylated filler DNA. Our data show high correlation between low input cfMBD-seq and standard MBD-seq (>1000 ng input). When compared to cfMEDIP-seq, cfMBD-seq demonstrates higher sequencing data quality with more sequenced reads passed filter and less duplicate rate. cfMBD-seq also outperforms cfMeDIP-seq in the enrichment of CpG islands. This new bisulphite-free ultra-low input methylation profiling technology has great potential in non-invasive and cost-effective cancer detection and classification.

Could Extracellular Vesicles Contribute to Generation or Awakening of "Sleepy" Metastatic Niches?

Pre-metastatic niches provide favorable conditions for tumor cells to disseminate, home to and grow in otherwise unfamiliar and distal microenvironments. Tumor-derived extracellular vesicles are now recognized as carriers of key messengers secreted by primary tumors, signals that induce the formation of pre-metastatic niches. Recent evidence suggests that tumor cells can disseminate from the very earliest stages of primary tumor development. However, once they reach distal sites, tumor cells can persist in a dormant state for long periods of time until their growth is reactivated and they produce metastatic lesions. In this new scenario, the question arises as to whether extracellular vesicles could influence the formation of these metastatic niches with dormant tumor cells? (here defined as "sleepy niches"). If so, what are the molecular mechanisms involved? In this perspective-review article, we discuss the possible influence of extracellular vesicles in early metastatic dissemination and whether they might play a role in tumor cell dormancy. In addition, we comment whether extracellular vesicle-mediated signals may be involved in tumor cell awakening, considering the possibility that extracellular vesicles might serve as biomarkers to detect early metastasis and/or minimal residual disease (MRD) monitoring.

Hypermethylation of Cyclin D2 Predicts Poor Prognosis of Hepatitis B Virus-Associated Hepatocellular Carcinoma after Hepatectomy

Prognosis of patients with hepatocellular carcinoma remains poor because of progression of hepatocellular carcinoma and high recurrence rates. Cyclin D2 (CCND2) plays a vital role in regulating the cell cycle; indeed, aberrant methylation of CCND2 is involved in the development of hepatocellular carcinoma. Therefore, we aimed to investigate levels of CCND2 methylation in patients with hepatitis B virus (HBV)-associated hepatocellular carcinoma and to evaluate its prognostic significance after hepatectomy. In total, 257 subjects were enrolled (166 hepatocellular carcinoma patients undergoing surgical resection, 61 chronic hepatitis B (CHB) patients, and 30 healthy controls). CCND2 methylation in peripheral blood mononuclear cells was measured quantitatively using MethyLight. We found that CCND2 methylation levels in patients with HBV-associated hepatocellular carcinoma were significantly higher than in CHB patients (P < 0.001) or healthy controls (P < 0.001). Within the hepatocellular carcinoma group, CCND2 methylation levels were higher in patients with portal vein invasion, early tumor recurrence, TNM III/IV stage, and tumor size ≥ 5 cm (P < 0.05). Furthermore, higher levels of CCND2 methylation were associated with worse overall survival and disease-free survival (P = 0.005 and P < 0.001, respectively). Multivariate analysis identified CCND2 methylation as an independent prognostic factor for early tumor recurrence (P = 0.021), overall survival (P = 0.022), and disease-free survival (P < 0.001) in hepatocellular carcinoma patients after resection. In conclusion, hypermethylation of CCND2 may have clinical utility for predicting a high risk of poor prognosis and early tumor recurrence in patients with HBV-associated hepatocellular carcinoma after hepatectomy.