Liquid Biopsy in Type 2 Diabetes Mellitus Management: Building Specific Biosignatures via Machine Learning

Liquid Biopsy: A Game Changer for Type 2 Diabetes

As the burden of type 2 diabetes (T2D) continues to escalate globally, there is a growing need for novel, less-invasive biomarkers capable of early diabetes detection and monitoring of disease progression. Liquid biopsy, recognized for its minimally invasive nature, is increasingly being applied beyond oncology, and nevertheless shows its potential when the collection of the tissue biopsy is not possible. This diagnostic approach involves utilizing liquid biopsy markers such as cell-free nucleic acids, extracellular vesicles, and diverse metabolites for the molecular diagnosis of T2D and its related complications. In this context, we thoroughly examine recent developments in T2D liquid biopsy research. Additionally, we discuss the primary challenges and future prospects of employing liquid biopsy in the management of T2D. Prognosis, diagnosis and monitoring of T2D through liquid biopsy could be a game-changing technique for personalized diabetes management.

A characteristic cerebellar biosignature for bipolar disorder, identified with fully automatic machine learning

Background

Transcriptomic profile differences between patients with bipolar disorder and healthy controls can be identified using machine learning and can provide information about the potential role of the cerebellum in the pathogenesis of bipolar disorder.With this aim, user-friendly, fully automated machine learning algorithms can achieve extremely high classification scores and disease-related predictive biosignature identification, in short time frames and scaled down to small datasets.

Method

A fully automated machine learning platform, based on the most suitable algorithm selection and relevant set of hyper-parameter values, was applied on a preprocessed transcriptomics dataset, in order to produce a model for biosignature selection and to classify subjects into groups of patients and controls. The parent GEO datasets were originally produced from the cerebellar and parietal lobe tissue of deceased bipolar patients and healthy controls, using Affymetrix Human Gene 1.0 ST Array.

Results

Patients and controls were classified into two separate groups, with no close-to-the-boundary cases, and this classification was based on the cerebellar transcriptomic biosignature of 25 features (genes), with Area Under Curve 0.929 and Average Precision 0.955. The biosignature includes both genes connected before to bipolar disorder, depression, psychosis or epilepsy, as well as genes not linked before with any psychiatric disease. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed participation of 4 identified features in 6 pathways which have also been associated with bipolar disorder.

Conclusion

Automated machine learning (AutoML) managed to identify accurately 25 genes that can jointly - in a multivariate-fashion - separate bipolar patients from healthy controls with high predictive power. The discovered features lead to new biological insights. Machine Learning (ML) analysis considers the features in combination (in contrast to standard differential expression analysis), removing both irrelevant as well as redundant markers, and thus, focusing to biological interpretation.

Automated machine learning for genome wide association studies

MOTIVATION

Genome-wide association studies (GWAS) present several computational and statistical challenges for their data analysis, including knowledge discovery, interpretability, and translation to clinical practice.

RESULTS

We develop, apply, and comparatively evaluate an automated machine learning (AutoML) approach, customized for genomic data that delivers reliable predictive and diagnostic models, the set of genetic variants that are important for predictions (called a biosignature), and an estimate of the out-of-sample predictive power. This AutoML approach discovers variants with higher predictive performance compared to standard GWAS methods, computes an individual risk prediction score, generalizes to new, unseen data, is shown to better differentiate causal variants from other highly correlated variants, and enhances knowledge discovery and interpretability by reporting multiple equivalent biosignatures.

AVAILABILITY AND IMPLEMENTATION

Code for this study is available at: https://github.com/mensxmachina/autoML-GWAS. JADBio offers a free version at: https://jadbio.com/sign-up/. SNP data can be downloaded from the EGA repository (https://ega-archive.org/). PRS data are found at: https://www.aicrowd.com/challenges/opensnp-height-prediction. Simulation data to study population structure can be found at: https://easygwas.ethz.ch/data/public/dataset/view/1/.

Regional differences in islet amyloid deposition in the residual pancreas with new-onset diabetes secondary to pancreatic ductal adenocarcinoma

BACKGROUND

Islet amyloid deposition and reduced β-cell mass are pathological hallmarks in type 2 diabetes mellitus subjects. To date, the pathological features of the islets in diabetes secondary to pancreatic ductal adenocarcinoma (PDAC) have not been specifically addressed.

AIM

To provide further insight into the relationship between islet amyloid deposition of the residual pancreas in PDAC patients and to explore whether regional differences (proximal vs distal residual pancreas) are associated with islet amyloid deposition.

METHODS

We retrospectively collected clinical information and pancreatic tissue removed from tumors of 45 PDAC patients, including 14 patients with normal glucose tolerance (NGT), 16 patients with prediabetes and 15 new-onset diabetes (NOD) patients diagnosed before surgery by an oral glucose tolerance test at West China Hospital from July 2017 to June 2020. Pancreatic volume was calculated by multiplying the estimated area of pancreatic tissue on each image slice by the interval between slices based on abdominal computer tomography scans. Several sections of paraffin-embedded pancreas specimens from both the proximal and/or distal regions remote from the tumor were stained as follows: (1) Hematoxylin and eosin for general histological appearance; (2) hematoxylin and insulin for the determination of fractional β-cell area (immunohistochemistry); and (3) quadruple insulin, glucagon, thioflavin T and DAPI staining for the determination of β-cell area, α-cell area and amyloid deposits.

RESULTS

Screening for pancreatic histologic features revealed that duct obstruction with islet amyloid deposition, fibrosis and marked acinar atrophy were robust in the distal pancreatic regions but much less robust in the proximal regions, especially in the prediabetes and NOD groups. Consistent with this finding, the remnant pancreatic volume was markedly decreased in the NOD group by nearly one-half compared with that in the NGT group (37.35 ± 12.16 cm3 vs 69.79 ± 18.17 cm3, P < 0.001). As expected, islets that stained positive for amyloid (islet amyloid density) were found in the majority of PDAC cases. The proportion of amyloid/islet area (severity of amyloid deposition) was significantly higher in both prediabetes and NOD patients than in NGT patients (P = 0.002; P < 0.0001, respectively). We further examined the regional differences in islet amyloid deposits. Islet amyloid deposit density was robustly increased by approximately 8-fold in the distal regions compared with that in the proximal regions in the prediabetes and NOD groups (3.98% ± 3.39% vs 0.50% ± 0.72%, P = 0.01; 12.03% vs 1.51%, P = 0.001, respectively).

CONCLUSION

In conclusion, these findings suggest that robust alterations of the distal pancreas due to tumors can disturb islet function and structure with islet amyloid formation, which may be associated with the pathogenesis of NOD secondary to PDAC.

Label-Free Human Disease Characterization through Circulating Cell-Free DNA Analysis Using Raman Spectroscopy

Circulating cell-free DNA (ccfDNA) is a liquid biopsy biomaterial attracting significant attention for the implementation of precision medicine diagnostics. Deeper knowledge related to its structure and biology would enable the development of such applications. In this study, we employed Raman spectroscopy to unravel the biomolecular profile of human ccfDNA in health and disease. We established reference Raman spectra of ccfDNA samples from healthy males and females with different conditions, including cancer and diabetes, extracting information about their chemical composition. Comparative observations showed a distinct spectral pattern in ccfDNA from breast cancer patients taking neoadjuvant therapy. Raman analysis of ccfDNA from healthy, prediabetic, and diabetic males uncovered some differences in their biomolecular fingerprints. We also studied ccfDNA released from human benign and cancer cell lines and compared it to their respective gDNA, confirming it mirrors its cellular origin. Overall, we explored for the first time Raman spectroscopy in the study of ccfDNA and provided spectra of samples from different sources. Our findings introduce Raman spectroscopy as a new approach to implementing liquid biopsy diagnostics worthy of further elaboration.

Identification of Metabolism-Related Proteins as Biomarkers of Insulin Resistance and Potential Mechanisms of m6A Modification

BACKGROUND

Insulin resistance (IR) is a major contributing factor to the pathogenesis of metabolic syndrome and type 2 diabetes mellitus (T2D). Adipocyte metabolism is known to play a crucial role in IR. Therefore, the aims of this study were to identify metabolism-related proteins that could be used as potential biomarkers of IR and to investigate the role of N⁶-methyladenosine (m⁶A) modification in the pathogenesis of this condition.

METHODS

RNA-seq data on human adipose tissue were retrieved from the Gene Expression Omnibus database. The differentially expressed genes of metabolism-related proteins (MP-DEGs) were screened using protein annotation databases. Biological function and pathway annotations of the MP-DEGs were performed through Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analyses. Key MP-DEGs were screened, and a protein-protein interaction (PPI) network was constructed using STRING, Cytoscape, MCODE, and CytoHubba. LASSO regression analysis was used to select primary hub genes, and their clinical performance was assessed using receiver operating characteristic (ROC) curves. The expression of key MP-DEGs and their relationship with m⁶A modification were further verified in adipose tissue samples collected from healthy individuals and patients with IR.

RESULTS

In total, 69 MP-DEGs were screened and annotated to be enriched in pathways related to hormone metabolism, low-density lipoprotein particle and carboxylic acid transmembrane transporter activity, insulin signaling, and AMPK signaling. The MP-DEG PPI network comprised 69 nodes and 72 edges, from which 10 hub genes (FASN, GCK, FGR, FBP1, GYS2, PNPLA3, MOGAT1, SLC27A2, PNPLA3, and ELOVL6) were identified. FASN was chosen as the key gene because it had the highest maximal clique centrality (MCC) score. GCK, FBP1, and FGR were selected as primary genes by LASSO analysis. According to the ROC curves, GCK, FBP1, FGR, and FASN could be used as potential biomarkers to detect IR with good sensitivity and accuracy (AUC = 0.80, 95% CI: 0.67-0.94; AUC = 0.86, 95% CI: 0.74-0.94; AUC = 0.83, 95% CI: 0.64-0.92; AUC = 0.78, 95% CI: 0.64-0.92). The expression of FASN, GCK, FBP1, and FGR was significantly correlated with that of IGF2BP3, FTO, EIF3A, WTAP, METTL16, and LRPPRC (p < 0.05). In validation clinical samples, the FASN was moderately effective for detecting IR (AUC = 0.78, 95% CI: 0.69-0.80), and its expression was positively correlated with the methylation levels of FASN (r = 0.359, p = 0.001).

CONCLUSION

Metabolism-related proteins play critical roles in IR. Moreover, FASN and GCK are potential biomarkers of IR and may be involved in the development of T2D via their m⁶A modification. These findings offer reliable biomarkers for the early detection of T2D and promising therapeutic targets.

Adipose cell-free DNA in diabetes

Cancer-associated necrosis is a well-known source of cell-free DNA (cfDNA). However, the origins of cfDNA are not strictly limited to cancer. Additionally, dietary exposure induces apoptosis-induced proliferation in adipocytes, leading to the release of cfDNA. The genetic information derived from cfDNA as a result of apoptosis-induced proliferation contains specific methylation patterns in adipose tissue that can be used as a marker to detect the risk of developing Type 2 diabetes Mellitus (T2DM) in the future. cfDNA is superior to peripheral blood leukocytes (PBL) and whole blood samples for reflecting tissue pathology due to the frequent use of PBL and whole blood samples that do not match tissue pathology. The difficulty of demonstrating that cfDNA is derived from adipose tissue. We propose several promising techniques by analyzing cfDNA derived from adipose tissue to detect T2DM risk. First, adipose-specific genes such as ADIPOQ and Leptin were utilized. Second, MCTA-Seq, EpiSCORE, deconvolution, multiplexing, and automated machine learning (AutoML) were used to determine the proportion of total methylation in related genes.