Novel Strategy for Human Deep Vein Thrombosis Diagnosis Based on Metabolomics and Stacking Machine Learning.

Deep vein thrombosis (DVT) is a serious health issue that often leads to considerable morbidity and mortality. Diagnosis of DVT in a clinical setting, however, presents considerable challenges. The fusion of metabolomics techniques and machine learning methods has led to high diagnostic and prognostic accuracy for various pathological conditions. This study explored the synergistic potential of dual-platform metabolomics (specifically, gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS)) to expand the detection of metabolites and improve the precision of DVT diagnosis. Sixty-one differential metabolites were identified in serum from DVT patients: 22 from GC-MS and 39 from LC-MS. Among these, five key metabolites were highlighted by SHapley Additive exPlanations (SHAP)-guided feature engineering and then used to develop a stacking diagnostic model. Additionally, a user-friendly interface application system was developed to streamline and automate the application of the diagnostic model, enhancing its practicality and accessibility for clinical use. This work showed that the integration of dual-platform metabolomics with a stacking machine learning model enables faster and more accurate diagnosis of DVT in clinical environments.

Combined metabolomics and tandem machine-learning models for wound age estimation: a novel analytical strategy.

Wound age estimation is one of the most challenging and indispensable issues for forensic pathologists. Although many methods based on physical findings and biochemical tests can be used to estimate wound age, an objective and reliable method for inferring the time interval after injury remains difficult. In the present study, endogenous metabolites of contused skeletal muscle were investigated to estimate the time interval after injury. Animal model of skeletal muscle injury was established using Sprague-Dawley rat, and the contused muscles were sampled at 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, and 48 h postcontusion (n = 9). Then, the samples were analysed using ultraperformance liquid chromatography coupled with high-resolution mass spectrometry. A total of 43 differential metabolites in contused muscle were determined by metabolomics method. They were applied to construct a two-level tandem prediction model for wound age estimation based on multilayer perceptron algorithm. As a result, all muscle samples were eventually divided into the following subgroups: 4, 8, 12, 16-20, 24-32, 36-40, and 44-48 h. The tandem model exhibited a robust performance and achieved a prediction accuracy of 92.6%, which was much higher than that of the single model. In summary, the multilayer perceptron-multilayer perceptron tandem machine-learning model based on metabolomics data can be used as a novel strategy for wound age estimation in future forensic casework. Key Points: The changes of metabolite profile were correlated with the time interval after injury in contused skeletal muscle.A panel of 43 endogenous metabolites screened by ultraperformance liquid chromatography coupled with high-resolution mass spectrometry could distinguish the wound ages.The multilayer perceptron (MLP) algorithm exhibited a robust performance in wound age estimation using metabolites.The combination of matabolomics and MLP-MLP tandem model could improve the accuracy of inferring the time interval after injury.

Exploring postmortem succession of rat intestinal microbiome for PMI based on machine learning algorithms and potential use for humans.

The microbial communities may undergo a meaningful successional change during the progress of decay and decomposition that could aid in determining the post-mortem interval (PMI). However, there are still challenges to applying microbiome-based evidence in law enforcement practice. In this study, we attempted to investigate the principles governing microbial community succession during decomposition of rat and human corpse, and explore their potential use for PMI of human cadavers. A controlled experiment was conducted to characterize temporal changes in microbial communities associated with rat corpses as they decomposed for 30 days. Obvious differences of microbial community structures were observed among different stages of decomposition, especially between decomposition of 0-7d and 9-30d. Thus, a two-layer model for PMI prediction was developed based on the succession of bacteria by combining classification and regression models using machine learning algorithms. Our results achieved 90.48% accuracy for discriminating groups of PMI 0-7d and 9-30d, and yielded a mean absolute error of 0.580d within 7d decomposition and 3.165d within 9-30d decomposition. Furthermore, samples from human cadavers were collected to gain the common succession of microbial community between rats and humans. Based on the 44 shared genera of rats and humans, a two-layer model of PMI was rebuilt to be applied for PMI prediction of human cadavers. Accurate estimates indicated a reproducible succession of gut microbes across rats and humans. Together these results suggest that microbial succession was predictable and can be developed into a forensic tool for estimating PMI.

Forensic identification of sudden cardiac death: a new approach combining metabolomics and machine learning.

The determination of sudden cardiac death (SCD) is one of the difficult tasks in the forensic practice, especially in the absence of specific morphological changes in the autopsies and histological investigations. In this study, we combined the metabolic characteristics from corpse specimens of cardiac blood and cardiac muscle to predict SCD. Firstly, ultra-high performance liquid chromatography coupled with high-resolution mass spectrometry (UPLC-HRMS)-based untargeted metabolomics was applied to obtain the metabolomic profiles of the specimens, and 18 and 16 differential metabolites were identified in the cardiac blood and cardiac muscle from the corpses of those who died of SCD, respectively. Several possible metabolic pathways were proposed to explain these metabolic alterations, including the metabolism of energy, amino acids, and lipids. Then, we validated the capability of these combinations of differential metabolites to distinguish between SCD and non-SCD through multiple machine learning algorithms. The results showed that stacking model integrated differential metabolites featured from the specimens showed the best performance with 92.31% accuracy, 93.08% precision, 92.31% recall, 91.96% F1 score, and 0.92 AUC. Our results revealed that the SCD metabolic signature identified by metabolomics and ensemble learning in cardiac blood and cardiac muscle has potential in SCD post-mortem diagnosis and metabolic mechanism investigations.

GPR65 as a potential immune checkpoint regulates the immune microenvironment according to pan-cancer analysis.

It has been reported that inhibition of GPR65 may be effective for the treatment of certain cancers. Nevertheless, the role of GPR65 in various cancers remains unknown. We conducted an exhaustive pan-cancer analysis of GPR65 using multiple databases, including TCGA, GTEx, BioGPS, HPA, cBioPortal, and GeneCards. GPR65 was found to be differentially expressed in various cancers and linked to tumor mutational burden (TMB), microsatellite instability (MSI), and Ploidy, playing a key function in the tumor microenvironment (TME). It is closely linked to the development of Th17 cells as well as Th1 and Th2 cells in certain cancers. Our findings indicate that the expression of GPR65 is highly linked with clinical prognosis, mutations, and immune cell infiltration. It was revealed as an indicator of patient prognosis as well as a possible immunomodulatory role. As a possible new immunological checkpoint, GPR65 could be a target for tumor immunotherapy.

Combined metabolomics and machine learning algorithms to explore metabolic biomarkers for diagnosis of acute myocardial ischemia.

Acute myocardial ischemia (AMI) remains the leading cause of death worldwide, and the post-mortem diagnosis of AMI represents a current challenge for both clinical and forensic pathologists. In the present study, the untargeted metabolomics based on ultra-performance liquid chromatography combined with high-resolution mass spectrometry was applied to analyze serum metabolic signatures from AMI in a rat model (n = 10 per group). A total of 28 endogenous metabolites in serum were significantly altered in AMI group relative to control and sham groups. A set of machine learning algorithms, namely gradient tree boosting (GTB), support vector machine (SVM), random forest (RF), logistic regression (LR), and multilayer perceptron (MLP) models, was used to screen the more valuable metabolites from 28 metabolites to optimize the biomarker panel. The results showed that classification accuracy and performance of MLP model were better than other algorithms when the metabolites consisting of L-threonic acid, N-acetyl-L-cysteine, CMPF, glycocholic acid, L-tyrosine, cholic acid, and glycoursodeoxycholic acid. Finally, 17 blood samples from autopsy cases were applied to validate the classification model's value in human samples. The MLP model constructed based on rat dataset achieved accuracy of 88.23%, and ROC of 0.89 for predicting AMI type II in autopsy cases of sudden cardiac death. The results demonstrated that MLP model based on 7 molecular biomarkers had a good diagnostic performance for both AMI rats and autopsy-based blood samples. Thus, the combination of metabolomics and machine learning algorithms provides a novel strategy for AMI diagnosis.

A novel method for determining postmortem interval based on the metabolomics of multiple organs combined with ensemble learning techniques.

Determining postmortem interval (PMI) is one of the most challenging and essential endeavors in forensic science. Developments in PMI estimation can take advantage of machine learning techniques. Currently, applying an algorithm to obtain information on multiple organs and conducting joint analysis to accurately estimate PMI are still in the early stages. This study aimed to establish a multi-organ stacking model that estimates PMI by analyzing differential compounds of four organs in rats. In a total of 140 rats, skeletal muscle, liver, lung, and kidney tissue samples were collected at each time point after death. Ultra-performance liquid chromatography coupled with high-resolution mass spectrometry was used to determine the compound profiles of the samples. The original data were preprocessed using multivariate statistical analysis to determine discriminant compounds. In addition, three interrelated and increasingly complex patterns (single organ optimal model, single organ stacking model, multi-organ stacking model) were established to estimate PMI. The accuracy and generalized area under the receiver operating characteristic curve of the multi-organ stacking model were the highest at 93% and 0.96, respectively. Only 1 of the 14 external validation samples was misclassified by the multi-organ stacking model. The results demonstrate that the application of the multi-organ combination to the stacking algorithm is a potential forensic tool for the accurate estimation of PMI.

Analysis of sensitivity and specificity: precise recognition of neutrophils during regeneration of contused skeletal muscle in rats.

In this report, we applied the TissueFAXS 200 digital pathological analysis system to rapidly and accurately identify neutrophils during regeneration of contused skeletal muscle, and to provide information for follow-up studies on neutrophils to estimate wound age. Rat injury model was established, and skeletal muscle samples were obtained from the control group and contusion groups at 1, 1.5, 2, 3, 4, and 6 h, as well as at 1, 3, 5, and 15 d post-injury (n = 5 per group). The expression of nuclei and neutrophils was detected by hematoxylin and eosin (HE) staining and immunohistochemical (IHC) staining. A total of 20 injury site areas of 0.25 mm2 (0.5 mm × 0.5 mm) were then randomly selected at all time points. A TissueFAXS 200 digital pathological analysis system was used to identify the positive and negative numbers. Knowledge of five professional medical workers were considered the gold standard to measure the false positive rate (FPR), false negative rate (FNR), sensitivity, specificity, and area under the curve (AUC) of receiver operating characteristic (ROC) curves. As a result, with a staining area of neutrophils from 8 µm2 to 15 µm2, the FPR was 4.28%-12.14%, the FNR was 12.42%-64.08%, the sensitivity was 35.92%-87.58%, the specificity was 87.86%-95.72%, the Youden index was 0.316-0.754, the accuracy was 82.80%-88.30%, and the AUC was 0.771-0.826. The AUC was largest when the cut-off value of the staining area was 12 µm2. Our results show that this software-based method is more accurate than the human eye in evaluating neutrophil infiltration. Based on the sensitivity and specificity, neutrophils can be accurately identified during regeneration of contused skeletal muscle. The TissueFAXS 200 digital pathological analysis system can also be used to optimize conditions for different cell types under various injury conditions to determine the optimal cut-off value of the staining area and provide optimal conditions for further study. Furthermore, it will provide evidence for forensic pathology cases.

Wound age estimation based on next-generation sequencing: Fitting the optimal index system using machine learning.

Accurate estimation of the wound age is critical in investigating intentional injury cases. Establishing objective and reliable biological indicators to estimate wound age is still a significant challenge in forensic medicine. Therefore, exploring an objective, flexible, and reliable index system selection method for wound age estimation based on next-generation sequencing gene expression profiles is necessary. We randomly divided 63 Sprague-Dawley rats into a control group, seven experimental groups (n = 7 per group), and an external validation group. After rats in the experimental and external validation groups suffered contusions, we sacrificed them at 4, 8, 12, 16, 20, 24, and 48 h after contusion, respectively. We selected 54 genes with the most significant changes between adjacent time points after contusion and defined set A. The Hub genes with time-related expression patterns were set B, C, and D through next-generation sequencing and bioinformatics analysis. Four different machine learning classification algorithms, including logistic regression, support vector machine, multi-layer perceptron, and random forest were used to compare and verify the efficiency of four index systems to estimate the wound age. The best combination for wound age estimation is the Genes ascribed to set A combined with the random forest classification algorithm. The accuracy of external verification was 85.71%. Only one rat was incorrectly classified (4 h post-injury incorrectly classified as 8 h). This study demonstrated the potential advantage of the index system selection based on next-generation sequencing and bioinformatics analysis for wound age estimation.

Investigating Transcriptional Dynamics Changes and Time-Dependent Marker Gene Expression in the Early Period After Skeletal Muscle Injury in Rats.

Following skeletal muscle injury (SMI), from post-injury reaction to repair consists of a complex series of dynamic changes. However, there is a paucity of research on detailed transcriptional dynamics and time-dependent marker gene expression in the early stages after SMI. In this study, skeletal muscle tissue in rats was taken at 4 to 48 h after injury for next-generation sequencing. We examined the transcriptional kinetics characteristics during above time periods after injury. STEM and maSigPro were used to screen time-correlated genes. Integrating 188 time-correlated genes with 161 genes in each time-related gene module by WGCNA, we finally identified 18 network-node regulatory genes after SMI. Histological staining analyses confirmed the mechanisms underlying changes in the tissue damage to repair process. Our research linked a variety of dynamic biological processes with specific time periods and provided insight into the characteristics of transcriptional dynamics, as well as screened time-related biological indicators with biological significance in the early stages after SMI.

Estimating Postmortem Interval Using Intestinal Microbiota Diversity Based on 16S rRNA High-throughput Sequencing Technology. / 基于16S rRNAé«˜é€šé‡æµ‹åºæŠ€æœ¯çš„è‚ é“èŒç¾¤å¤šæ ·æ€§æŽ¨æ–­æ­»äº¡æ—¶é—´.

OBJECTIVES: To explore the correlation between intestinal microbiota and postmortem interval(PMI) in rats by using 16S rRNA high-throughput sequencing technology. METHODS: Rats were killed by anesthesia and placed at 16 ℃, and DNA was extracted in caecum at 14 time points of 0, 1, 2, 3, 5, 7, 9, 12, 15, 18, 21, 24, 27 and 30 d after death. The 16S rRNA high-throughput sequencing technology was used to detect intestinal microbiota in rat cecal contents, and the results were used to analyze the rat intestinal microbiota diversity and differences. RESULTS: The total number of intestinal microbial communities did not change significantly within 30 days after death, but the diversity showed an upward trend. A total of 119 bacterial communities were significantly changed at 13 time points after death. The models for PMI estimation were established by using partial least squares (PLS) regression at all time points, before 9 days and after 12 days, reaching an R2 of 0.795, 0.767 and 0.445, respectively; and the root mean square errors (RMSEs) were 6.57, 1.96 and 5.37 d, respectively. CONCLUSIONS: Using 16S rRNA high-throughput sequencing technology, the composition and structure of intestinal microbiota changed significantly within 30 d after death. In addition, the established PLS regression model suggested that the PMI was highly correlated with intestinal microbiota composition, showing a certain time series change.

Novel insights into wound age estimation: combined with "up, no change, or down" system and cosine similarity in python environment.

Wound age estimation is a complex, multifactorial issue. It is considered to have great practical significance that combining multi-biomarkers and multi-methods for injury time estimation. We optimized our earlier "up, no change, or down" model by adding data on the expression levels of mRNAs encoding ABHD2, MAD2L2, and ARID5A, and we converted the relative quantitative expression levels of seven genes into a vector rather than a color model. We used Python to derive the cosine similarity (CS) between a test set and the vector matrix; the highest similarity most accurately reflected the injury time. For the optimized model, the internal and external verifications were approximately 0.71 and 0.66, respectively. The good double-blinded results indicated that the model was stable and reliable. In summary, we used a vector matrix and cosine similarities derived by Python to mine the levels of genes expressed in contused skeletal muscle. We are the first to combine several biomarkers and methods for wound age estimation.