EfficientNet-based machine learning architecture for sleep apnea identification in clinical single-lead ECG signal data sets.

OBJECTIVE: Our objective was to create a machine learning architecture capable of identifying obstructive sleep apnea (OSA) patterns in single-lead electrocardiography (ECG) signals, exhibiting exceptional performance when utilized in clinical data sets. METHODS: We conducted our research using a data set consisting of 1656 patients, representing a diverse demographic, from the sleep center of China Medical University Hospital. To detect apnea ECG segments and extract apnea features, we utilized the EfficientNet and some of its layers, respectively. Furthermore, we compared various training and data preprocessing techniques to enhance the model's prediction, such as setting class and sample weights or employing overlapping and regular slicing. Finally, we tested our approach against other literature on the Apnea-ECG database. RESULTS: Our research found that the EfficientNet model achieved the best apnea segment detection using overlapping slicing and sample-weight settings, with an AUC of 0.917 and an accuracy of 0.855. For patient screening with AHI > 30, we combined the trained model with XGBoost, leading to an AUC of 0.975 and an accuracy of 0.928. Additional tests using PhysioNet data showed that our model is comparable in performance to existing models regarding its ability to screen OSA levels. CONCLUSIONS: Our suggested architecture, coupled with training and preprocessing techniques, showed admirable performance with a diverse demographic dataset, bringing us closer to practical implementation in OSA diagnosis. Trial registration The data for this study were collected retrospectively from the China Medical University Hospital in Taiwan with approval from the institutional review board CMUH109-REC3-018.

Multi-omics profiling reveals microRNA-mediated insulin signaling networks.

BACKGROUND: MicroRNAs (miRNAs) play a key role in mediating the action of insulin on cell growth and the development of diabetes. However, few studies have been conducted to provide a comprehensive overview of the miRNA-mediated signaling network in response to glucose in pancreatic beta cells. In our study, we established a computational framework integrating multi-omics profiles analyses, including RNA sequencing (RNA-seq) and small RNA sequencing (sRNA-seq) data analysis, inverse expression pattern analysis, public data integration, and miRNA targets prediction to illustrate the miRNA-mediated regulatory network at different glucose concentrations in INS-1 pancreatic beta cells (INS-1), which display important characteristics of the pancreatic beta cells. RESULTS: We applied our computational framework to the expression profiles of miRNA/mRNA of INS-1, at different glucose concentrations. A total of 1437 differentially expressed genes (DEGs) and 153 differentially expressed miRNAs (DEmiRs) were identified from multi-omics profiles. In particular, 121 DEmiRs putatively regulated a total of 237 DEGs involved in glucose metabolism, fatty acid oxidation, ion channels, exocytosis, homeostasis, and insulin gene regulation. Moreover, Argonaute 2 immunoprecipitation sequencing, qRT-PCR, and luciferase assay identified Crem, Fn1, and Stc1 are direct targets of miR-146b and elucidated that miR-146b acted as a potential regulator and promising target to understand the insulin signaling network. CONCLUSIONS: In this study, the integration of experimentally verified data with system biology framework extracts the miRNA network for exploring potential insulin-associated miRNA and their target genes. The findings offer a potentially significant effect on the understanding of miRNA-mediated insulin signaling network in the development and progression of pancreatic diabetes.

miRTarBase update 2018: a resource for experimentally validated microRNA-target interactions.

MicroRNAs (miRNAs) are small non-coding RNAs of ∼ 22 nucleotides that are involved in negative regulation of mRNA at the post-transcriptional level. Previously, we developed miRTarBase which provides information about experimentally validated miRNA-target interactions (MTIs). Here, we describe an updated database containing 422 517 curated MTIs from 4076 miRNAs and 23 054 target genes collected from over 8500 articles. The number of MTIs curated by strong evidence has increased ∼1.4-fold since the last update in 2016. In this updated version, target sites validated by reporter assay that are available in the literature can be downloaded. The target site sequence can extract new features for analysis via a machine learning approach which can help to evaluate the performance of miRNA-target prediction tools. Furthermore, different ways of browsing enhance user browsing specific MTIs. With these improvements, miRTarBase serves as more comprehensively annotated, experimentally validated miRNA-target interactions databases in the field of miRNA related research. miRTarBase is available at http://miRTarBase.mbc.nctu.edu.tw/.

Secretory Production of Functional Grouper Type I Interferon from Epinephelus septemfasciatus in Escherichia coli and Bacillus subtilis.

Nervous necrosis virus (NNV) results in high mortality rates of infected marine fish worldwide. Interferons (IFNs) are cytokines in vertebrates that suppress viral replication and regulate immune responses. Heterologous overexpression of fish IFN in bacteria could be problematic because of protein solubility and loss of function due to protein misfolding. In this study, a protein model of the IFN-α of Epinephelus septemfasciatus was built based on comparative modeling. In addition, PelB and SacB signal peptides were fused to the N-terminus of E. septemfasciatus IFN-α for overexpression of soluble, secreted IFN in Escherichia coli (E-IFN) and Bacillus subtilis (B-IFN). Cytotoxicity tests indicated that neither recombinant grouper IFN-α were cytotoxic to a grouper head kidney cell line (GK). The GK cells stimulated with E-IFN and B-IFN exhibited elevated expression of antiviral Mx genes when compared with the control group. The NNV challenge experiments demonstrated that GK cells pretreated or co-treated with E-IFN and B-IFN individually had three times the cell survival rates of untreated cells, indicating the cytoprotective ability of our recombinant IFNs. These data provide a protocol for the production of soluble, secreted, and functional grouper IFN of high purity, which may be applied to aquaculture fisheries for antiviral infection.

Deep learning assists in acute leukemia detection and cell classification via flow cytometry using the acute leukemia orientation tube.

This study aimed to evaluate the sensitivity of AI in screening acute leukemia and its capability to classify either physiological or pathological cells. Utilizing an acute leukemia orientation tube (ALOT), one of the protocols of Euroflow, flow cytometry efficiently identifies various forms of acute leukemia. However, the analysis of flow cytometry can be time-consuming work. This retrospective study included 241 patients who underwent flow cytometry examination using ALOT between 2017 and 2022. The collected flow cytometry data were used to train an artificial intelligence using deep learning. The trained AI demonstrated a 94.6% sensitivity in detecting acute myeloid leukemia (AML) patients and a 98.2% sensitivity for B-lymphoblastic leukemia (B-ALL) patients. The sensitivities of physiological cells were at least 80%, with variable performance for pathological cells. In conclusion, the AI, trained with ResNet-50 and EverFlow, shows promising results in identifying patients with AML and B-ALL, as well as classifying physiological cells.

Comparative analysis of machine learning algorithms for Alzheimer's disease classification using EEG signals and genetic information.

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline, memory impairments, and behavioral changes. The presence of abnormal beta-amyloid plaques and tau protein tangles in the brain is known to be associated with AD. However, current limitations of imaging technology hinder the direct detection of these substances. Consequently, researchers are exploring alternative approaches, such as indirect assessments involving monitoring brain signals, cognitive decline levels, and blood biomarkers. Recent studies have highlighted the potential of integrating genetic information into these approaches to enhance early detection and diagnosis, offering a more comprehensive understanding of AD pathology beyond the constraints of existing imaging methods. Our study utilized electroencephalography (EEG) signals, genotypes, and polygenic risk scores (PRSs) as features for machine learning models. We compared the performance of gradient boosting (XGB), random forest (RF), and support vector machine (SVM) to determine the optimal model. Statistical analysis revealed significant correlations between EEG signals and clinical manifestations, demonstrating the ability to distinguish the complexity of AD from other diseases by using genetic information. By integrating EEG with genetic data in an SVM model, we achieved exceptional classification performance, with an accuracy of 0.920 and an area under the curve of 0.916. This study presents a novel approach of utilizing real-time EEG data and genetic background information for multimodal machine learning. The experimental results validate the effectiveness of this concept, providing deeper insights into the actual condition of patients with AD and overcoming the limitations associated with single-oriented data.

Utility of polygenic scores across diverse diseases in a hospital cohort for predictive modeling.

Polygenic scores estimate genetic susceptibility to diseases. We systematically calculated polygenic scores across 457 phenotypes using genotyping array data from China Medical University Hospital. Logistic regression models assessed polygenic scores' ability to predict disease traits. The polygenic score model with the highest accuracy, based on maximal area under the receiver operating characteristic curve (AUC), is provided on the GeneAnaBase website of the hospital. Our findings indicate 49 phenotypes with AUC greater than 0.6, predominantly linked to endocrine and metabolic diseases. Notably, hyperplasia of the prostate exhibited the highest disease prediction ability (P value = 1.01 × 10-19, AUC = 0.874), highlighting the potential of these polygenic scores in preventive medicine and diagnosis. This study offers a comprehensive evaluation of polygenic scores performance across diverse human traits, identifying promising applications for precision medicine and personalized healthcare, thereby inspiring further research and development in this field.

Machine learning approaches for biomarker discovery to predict large-artery atherosclerosis.

Large-artery atherosclerosis (LAA) is a leading cause of cerebrovascular disease. However, LAA diagnosis is costly and needs professional identification. Many metabolites have been identified as biomarkers of specific traits. However, there are inconsistent findings regarding suitable biomarkers for the prediction of LAA. In this study, we propose a new method integrates multiple machine learning algorithms and feature selection method to handle multidimensional data. Among the six machine learning models, logistic regression (LR) model exhibited the best prediction performance. The value of area under the receiver operating characteristic curve (AUC) was 0.92 when 62 features were incorporated in the external validation set for the LR model. In this model, LAA could be well predicted by clinical risk factors including body mass index, smoking, and medications for controlling diabetes, hypertension, and hyperlipidemia as well as metabolites involved in aminoacyl-tRNA biosynthesis and lipid metabolism. In addition, we found that 27 features were present among the five adopted models that could provide good results. If these 27 features were used in the LR model, an AUC value of 0.93 could be achieved. Our study has demonstrated the effectiveness of combining machine learning algorithms with recursive feature elimination and cross-validation methods for biomarker identification. Moreover, we have shown that using shared features can yield more reliable correlations than either model, which can be valuable for future identification of LAA.

Structural insights into EphA4 unconventional activation from prediction of the EphA4 and its complex with ribonuclease 1.

It has been shown that several ribonuclease (RNase) A superfamily proteins serve as ligands of receptor tyrosine kinases (RTKs), representing a new concept for ligand/receptor interaction. Moreover, recent studies indicate high clinical values for this type of ligand/RTK interactions. However, there is no structural report for this new family of ligand/receptor. In an attempt to understand how RNase and RTK may interact, we focused on the RNase1/ephrin type-A receptor 4 (EphA4) complex and predicted their structure by using the state-of-the-art machine learning method, AlphaFold and its derivative method, AF2Complex. In this model, electrostatic force plays an essential role for the specific ligand/receptor interaction. We found the R39 of RNase1 is the key residue for EphA4-binding and activation. Mutation on this residue causes disruption of an essential basic patch, resulting in weaker ligand-receptor association and leading to the loss of activation. By comparing the surface charge distribution of the RNase A superfamily, we found the positively charged residues on the RNase1 surface is more accessible for EphA4 forming salt bridges than other RNases. Furthermore, RNase1 binds to the ligand-binding domain (LBD) of EphA4, which is responsible for the traditional ligand ephrin-binding. Our model reveals the location of RNase1 on EphA4 partially overlaps with that of ephrin-A5, a traditional ligand of EphA4, suggesting steric hindrance as the basis by which the ephrin-A5 precludes interactions of RNase1 with EphA4. Together, our discovery of RNase1/EphA4 interface provides a potential treatment strategy by blocking the RNase1-EphA4 axis.

Utilizing apolipoprotein E genotypes and associated comorbidities for the assessment of the risk for dementia.

Introduction: Dementia is associated with many comorbidities while being related to Apolipoprotein E (ApoE) polymorphism. However, it is unclear how these clinical illnesses and genetic factors modify the dementia risk. Methods: We enrolled 600 dementia cases and 6000 matched non-dementia controls, with identified ApoE genotype (ε4/ε4, ε4/ε3, and ε3/ε3). Eight comorbidities were selected by medical records, and counted if occurring within 3 years of enrollment. Results: The dementia group had a higher ratio of carrying ε4 allele and prevalence of comorbidities than the non-dementia group. Homozygous ε4 carriers presented the broken line of dementia risk with the peak age at 65-75 years and odds ratio (OR) up to 6.6. The risk only emerged after 65 years of age in ε3/ε4 subjects with OR around 1.6-2.4 when aged > 75 years. Cerebrovascular accident (CVA) is the commonest comorbidity (14.6%). CVA, sleep disorder, and functional gastrointestinal disorders remained as significant risk comorbidities for dementia throughout all age groups (OR = 1.7-5.0). When functional gastrointestinal disorder and ε4 allele both occurred, the dementia risk exceeded the summation of individual risks (OR = 3.7 and 1.9 individually, OR = 6.0 for the combination). Comorbidities could also be predictors of dementia. Conclusion: Combining the genetic and clinical information, we detected cognitive decline and optimize interventions early when the patients present a specific illness in a particular age and carry a specific ApoE allele. Of comorbidities, functional gastrointestinal disorder is the strongest predicting factor for dementia in ε4 allele carriers.

A Novel Quality-Control Procedure to Improve the Accuracy of Rare Variant Calling in SNP Arrays.

Background: Single-nucleotide polymorphism (SNP) arrays are an ideal technology for genotyping genetic variants in mass screening. However, using SNP arrays to detect rare variants [with a minor allele frequency (MAF) of <1%] is still a challenge because of noise signals and batch effects. An approach that improves the genotyping quality is needed for clinical applications. Methods: We developed a quality-control procedure for rare variants which integrates different algorithms, filters, and experiments to increase the accuracy of variant calling. Using data from the TWB 2.0 custom Axiom array, we adopted an advanced normalization adjustment to prevent false calls caused by splitting the cluster and a rare het adjustment which decreases false calls in rare variants. The concordance of allelic frequencies from array data was compared to those from sequencing datasets of Taiwanese. Finally, genotyping results were used to detect familial hypercholesterolemia (FH), thrombophilia (TH), and maturity-onset diabetes of the young (MODY) to assess the performance in disease screening. All heterozygous calls were verified by Sanger sequencing or qPCR. The positive predictive value (PPV) of each step was estimated to evaluate the performance of our procedure. Results: We analyzed SNP array data from 43,433 individuals, which interrogated 267,247 rare variants. The advanced normalization and rare het adjustment methods adjusted genotyping calling of 168,134 variants (96.49%). We further removed 3916 probesets which were discordant in MAFs between the SNP array and sequencing data. The PPV for detecting pathogenic variants with 0.01%10,000 are available. The results demonstrated our procedure could perform correct genotype calling of rare variants. It provides a solution of pathogenic variant detection through SNP array. The approach brings tremendous promise for implementing precision medicine in medical practice.

Enterotype-based Analysis of Gut Microbiota along the Conventional Adenoma-Carcinoma Colorectal Cancer Pathway.

The dysbiosis of human gut microbiota is strongly associated with the development of colorectal cancer (CRC). The dysbiotic features of the transition from advanced polyp to early-stage CRC are largely unknown. We performed a 16S rRNA gene sequencing and enterotype-based gut microbiota analysis study. In addition to Bacteroides- and Prevotella-dominated enterotypes, we identified an Escherichia-dominated enterotype. We found that the dysbiotic features of CRC were dissimilar in overall samples and especially Escherichia-dominated enterotype. Besides a higher abundance of Fusobacterium, Enterococcus, and Aeromonas in all CRC faecal microbiota, we found that the most notable characteristic of CRC faecal microbiota was a decreased abundance of potential beneficial butyrate-producing bacteria. Notably, Oscillospira was depleted in the transition from advanced adenoma to stage 0 CRC, whereas Haemophilus was depleted in the transition from stage 0 to early-stage CRC. We further identified 7 different CAGs by analysing bacterial clusters. The abundance of microbiota in cluster 3 significantly increased in the CRC group, whereas that of cluster 5 decreased. The abundance of both cluster 5 and cluster 7 decreased in the Escherichia-dominated enterotype of the CRC group. We present the first enterotype-based faecal microbiota analysis. The gut microbiota of colorectal neoplasms can be influenced by its enterotype.