Environmental chemical exposure dynamics and machine learning-based prediction of diabetes mellitus

The biomedical applications of artificial intelligence: an overview of decades of research

A significant area of computer science called artificial intelligence (AI) is successfully applied to the analysis of intricate biological data and the extraction of substantial associations from datasets for a variety of biomedical uses. AI has attracted significant interest in biomedical research due to its features: (i) better patient care through early diagnosis and detection; (ii) enhanced workflow; (iii) lowering medical errors; (v) lowering medical costs; (vi) reducing morbidity and mortality; (vii) enhancing performance; (viii) enhancing precision; and (ix) time efficiency. Quantitative metrics are crucial for evaluating AI implementations, providing insights, enabling informed decisions, and measuring the impact of AI-driven initiatives, thereby enhancing transparency, accountability, and overall impact. The implementation of AI in biomedical fields faces challenges such as ethical and privacy concerns, lack of awareness, technology unreliability, and professional liability. A brief discussion is given of the AI techniques, which include Virtual screening (VS), DL, ML, Hidden Markov models (HMMs), Neural networks (NNs), Generative models (GMs), Molecular dynamics (MD), and Structure-activity relationship (SAR) models. The study explores the application of AI in biomedical fields, highlighting its enhanced predictive accuracy, treatment efficacy, diagnostic efficiency, faster decision-making, personalised treatment strategies, and precise medical interventions.

Construction of risk prediction model of sentinel lymph node metastasis in breast cancer patients based on machine learning algorithm

PURPOSE

The aim of this study was to develop and validate a machine learning (ML) based prediction model for sentinel lymph node metastasis in breast cancer to identify patients with a high risk of sentinel lymph node metastasis.

METHODS

In this machine learning study, we retrospectively collected 225 female breast cancer patients who underwent sentinel lymph node biopsy (SLNB). Feature screening was performed using the logistic regression analysis. Subsequently, five ML algorithms, namely LOGIT, LASSO, XGBOOST, RANDOM FOREST model and GBM model were employed to train and develop an ML model. In addition, model interpretation was performed by the Shapley Additive Explanations (SHAP) analysis to clarify the importance of each feature of the model and its decision basis.

RESULTS

Combined univariate and multivariate logistic regression analysis, identified Multifocal, LVI, Maximum Diameter, Shape US, Maximum Cortical Thickness as significant predictors. We than successfully leveraged machine learning algorithms, particularly the RANDOM FOREST model, to develop a predictive model for sentinel lymph node metastasis in breast cancer. Finally, the SHAP method identified Maximum Diameter and Maximum Cortical Thickness as the primary decision factors influencing the ML model's predictions.

CONCLUSION

With the integration of pathological and imaging characteristics, ML algorithm can accurately predict sentinel lymph node metastasis in breast cancer patients. The RANDOM FOREST model showed ideal performance. With the incorporation of these models in the clinic, can helpful for clinicians to identify patients at risk of sentinel lymph node metastasis of breast cancer and make more reasonable treatment decisions.

Network analysis of core factors related to non-suicidal self-injury in adolescents with mood disorders

Introduction

Adolescents with mood disorders are at an exceptionally high risk for non-suicidal self-injury (NSSI); however, the understanding of the core factors underlying this vulnerability remains limited. This knowledge gap significantly hampers the effectiveness of targeted prevention and intervention strategies.

Methods

A total of 263 adolescents with mood disorders completed a series of self-report surveys, covering demographic, personal, and social factors related to NSSI. We first used least absolute shrinkage and selection operator (LASSO) regression to identify the core related factors. Then, we employed network analysis to construct the network structure of these core factors.

Results

Our findings indicate that depressive and anxiety symptoms are the strongest influencing factors for NSSI among adolescents with mood disorders. Life events and the specific functions of NSSI are identified as personalized factors within this group. Additionally, objective social support and education level emerged as potential protective factors against NSSI. These factors are not independent but interact with each other.

Conclusion

By identifying and intervening in these key factors, more effective prevention strategies and personalized treatment plans can be developed, ultimately improving the quality of life and psychological well-being of adolescents with mood disorders.

Improved detection of decreased glucose handling capacities via continuous glucose monitoring-derived indices

BACKGROUND

Efficiently assessing glucose handling capacity is a critical public health challenge. This study assessed the utility of relatively easy-to-measure continuous glucose monitoring (CGM)-derived indices in estimating glucose handling capacities calculated from resource-intensive clamp tests.

METHODS

We conducted a prospective study of 64 individuals without prior diabetes diagnosis. The study performed CGM, oral glucose tolerance tests (OGTT), and hyperglycemic and hyperinsulinemic-euglycemic clamp tests. We validated CGM-derived indices characteristics using an independent dataset from another country and mathematical models with simulated data.

RESULTS

A CGM-derived index reflecting the autocorrelation function of glucose levels (AC_Var) is significantly correlated with clamp-derived disposition index (DI), a well-established measure of glucose handling capacity and predictor of diabetes onset. Multivariate and machine learning models indicate AC_Var's contribution to predicting clamp-derived DI independent from other CGM-derived indices. The model using CGM-measured glucose standard deviation and AC_Var outperforms models using commonly used diabetes diagnostic indices, such as fasting blood glucose, HbA1c, and OGTT measures, in predicting clamp-derived DI. Mathematical simulations also demonstrate the association of AC_Var with DI.

CONCLUSIONS

CGM-derived indices, including AC_Var, serve as valuable tools for predicting glucose handling capacities in populations without prior diabetes diagnosis. We develop a web application that calculates these CGM-derived indices ( https://cgm-ac-mean-std.streamlit.app/ ).

Mechanistic investigation of Shuanghuanglian against infectious bronchitis in chickens: a network pharmacology and molecular dynamics study

Introduction

Infectious bronchitis (IB) poses a major challenge to global poultry production, causing substantial economic burdens and underscoring the necessity for novel therapeutic interventions given the limitations of current vaccines and conventional antiviral agents. The purpose of this study is to comprehensively explore the active components in Shuanghuanglian and their interaction with the key pathological targets of IBV (Infectious bronchitis virus) infection. By using advanced computational methods, this study aims not only to identify the therapeutic potential of active ingredients, but also to reveal their mechanism of action against IBV.

Methods

Through integrative systems pharmacology approaches, we systematically investigated Shuanghuanglian and its phytochemical constituents against IB, employing multi-omics analysis, ensemble machine learning, and all-atom molecular dynamics (MD) simulations. Network pharmacology revealed 65 target genes associated with Shuanghuanglian's primary bioactive components (quercetin, kaempferol, wogonin, and luteolin), exhibiting high network centrality.

Results

Using the TCMSP database, we found 65 target genes associated with key active components, such as quercetin and kaempferol, which exhibited strong connectivity in our network analysis. The GeneCards database also identified 40 common target genes shared by Shuanghuanglian and IB. Importantly, BCL2 and IL6 were recognized as key targets in the protein-protein interaction (PPI) network analysis, highlighting their roles in apoptosis and inflammation. Furthermore, analyses using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways revealed significant roles in regulating the cell cycle and inflammatory responses. Machine learning techniques identified BCL2 and IL6 as critical genes for therapeutic intervention, supported by molecular docking results that showed strong binding energies. Furthermore, molecular dynamics simulations confirm the stability of the complexes, underscoring the importance of these interactions for treatment efficacy.

Conclusion

We used a variety of analytical methods, and finally identified the potential active ingredients of Shuanghuanglian as kaempferol, quercetin, wogonin, and luteolin. The active ingredients target BCL2 and IL6 and play a therapeutic role in avian infectious bronchitis by inhibiting apoptosis and reducing inflammatory response.

Unveiling the effect of urinary xenoestrogens on chronic kidney disease in adults: A machine learning model

Exposure to three primary xenoestrogens (XEs), including phthalates, parabens, and phenols, has been strongly associated with chronic kidney disease (CKD). An interpretable machine learning (ML) model was developed to predict CKD using data from the National Health and Nutrition Examination Survey (NHANES) database spanning from 2007 to 2016. Four ML algorithms-random forest classifier (RF), XGBoost (XGB), k-nearest neighbors (KNN), and support vector machine (SVM)-were used alongside traditional logistic regression to predict CKD. The study included 6910 U.S. adults, with XGB showing the highest predictive accuracy, achieving an area under the curve (AUC) of 0.817 (95 % CI: 0.789, 0.844). The selected model was interpreted using Shapley additive explanations (SHAP) and partial dependence plot (PDP). The SHAP method identified key predictive features for CKD in urinary metabolites of XEs-methyl paraben (MeP), mono-(carboxynonyl) phthalate (MCNP), and triclosan (TCS)-and suggested personalized CKD care should focus on XE control. PDP results confirmed that, within certain ranges, MeP levels positively impacted the model, MCNP levels negatively impacted it, and TCS had a mixed effect. The synergistic effects suggested that managing urinary MeP levels could be essential for the effective control of CKD. In summary, our research highlights the significant predictive potential of XEs for CKD, especially MeP, MCNP, and TCS.

Development and validation of predictive models for diabetic retinopathy using machine learning

OBJECTIVE

This study aimed to develop and compare machine learning models for predicting diabetic retinopathy (DR) using clinical and biochemical data, specifically logistic regression, random forest, XGBoost, and neural networks.

METHODS

A dataset of 3,000 diabetic patients, including 1,500 with DR, was obtained from the National Population Health Science Data Center. Significant predictors were identified, and four predictive models were developed. Model performance was assessed using accuracy, precision, recall, F1-score, and area under the curve (AUC).

RESULTS

Random forest and XGBoost demonstrated superior performance, achieving accuracies of 95.67% and 94.67%, respectively, with AUC values of 0.991 and 0.989. Logistic regression yielded an accuracy of 76.50% (AUC: 0.828), while neural networks achieved 82.67% accuracy (AUC: 0.927). Key predictors included 24-hour urinary microalbumin, HbA1c, and serum creatinine.

CONCLUSION

The study highlights random forest and XGBoost as effective tools for early DR detection, emphasizing the importance of renal and glycemic markers in risk assessment. These findings support the integration of machine learning models into clinical decision-making for improved patient outcomes in diabetes management.

Construction and Verification of a Frailty Risk Prediction Model for Elderly Patients with Coronary Heart Disease Based on a Machine Learning Algorithm

Background

This study aimed to develop a machine learning-based predictive model for assessing frailty risk among elderly patients with coronary heart disease (CHD).

Methods

From November 2020 to May 2023, a cohort of 1170 elderly patients diagnosed with CHD were enrolled from the Department of Cardiology of a tier-3 hospital in Anhui Province, China. Participants were randomly divided into a development group and a validation group, each containing 585 patients in a 1:1 ratio. Least absolute shrinkage and selection operator (LASSO) regression was employed in the development group to identify key variables influencing frailty among patients with CHD. These variables informed the creation of a machine learning prediction model, with the most accurate model selected. Predictive accuracy was subsequently evaluated in the validation group through receiver operating characteristic (ROC) curve analysis.

Results

LASSO regression identified the activities of daily living (ADL) score, hemoglobin, low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC), depression, cardiac function classification, cerebrovascular disease, diabetes, solitary living, and age as significant predictors of frailty among elderly patients with CHD in the development group. These variables were incorporated into a logistic regression model and four machine learning models: extreme gradient boosting (XGBoost), random forest (RF), light gradient boosting machine (LightGBM), and adaptive boosting (AdaBoost). AdaBoost demonstrated the highest accuracy in the development group, achieving an area under the ROC curve (AUC) of 0.803 in the validation group, indicating strong predictive capability.

Conclusions

By leveraging key frailty determinants in elderly patients with CHD, the AdaBoost machine learning model developed in this study has shown robust predictive performance through validated indicators and offers a reliable tool for assessing frailty risk in this patient population.

Nomogram Based on HRV for Predicting the Therapeutic Effects of Orthostatic Training in Children with Vasovagal Syncope

BACKGROUND

This study intended to find out whether the parameters of heart rate variability (HRV) can predict the treatment efficacy of orthostatic training among pediatric cases of vasovagal syncope (VVS).

METHODS

Patients with VVS who underwent orthostatic training were retrospectively enrolled. Lasso and logistic regression were used to sift through variables and build the model, which is visualized using a nomogram. The model's performance was evaluated through calibration plots, a receiver operating characteristic (ROC) curve, and decision curve analysis (DCA) for both datasets.

RESULTS

In total, 119 participants were included in the analysis, and 73 and 46 were assigned to the training and validation datasets, respectively. Five factors with nonzero coefficients were chosen based on lasso regression: age, the root means square of successive differences between normal sinus beats (rMSSD), standard deviation of the averages normal-to-normal intervals in all 5-min segments, minimum heart rate, and high frequency. Drawing from the logistic regression analysis results, the visual predictive model incorporated two variables, namely age and rMSSD. For the training dataset, the sensitivity was 0.686 and the specificity was 0.868 with an area under the curve (AUC) of 0.81 (95% CI, 0.71-0.91) for the ROC curve. For the validation dataset, the AUC of the ROC was 0.80 (95% CI, 0.66-0.93), while sensitivity and specificity were recorded at 0.625 and 0.909, respectively. In the calibration plots for both datasets, the predicted probabilities correlated well with the actual probabilities. According to the DCA, the visual predictive model gained a significant net benefit across a wide threshold range.

CONCLUSIONS

Pediatric patients with VVS can benefit from orthostatic training using a visual predictive model comprising age and rMSSD.

Trends in the prevalence of osteoporosis and effects of heavy metal exposure using interpretable machine learning

There is limited evidence that heavy metals exposure contributes to osteoporosis. Multi-parameter scoring machine learning (ML) techniques were developed using National Health and Nutrition Examination Survey data to predict osteoporosis based on heavy metal exposure levels. For generating an optimal predictive model for osteoporosis, 12 ML models were used. Identification was carried out using the model that performed the best. For interpretation of models, Shapley additive explanation (SHAP) methods and partial dependence plots (PDP) were integrated into a pipeline and incorporated into the ML pipeline. By regressing osteoporosis on survey cycles, logistic regression was used to evaluate linear trends in osteoporosis over time. For the purpose of training and validating predictive models, 5745 eligible participants were randomly selected into training and testing set. It was evident from the results that the gradient boosting decision tree model performed the best among the predictive models, attributing to an accuracy rate of 89.40 % in the testing set. Based on the model results, the area under the curve and F1 score were 0.88 and 0.39, respectively. As a result of the SHAP analysis, urinary Co, urinary Tu, blood Cd, and urinary Hg levels were identified as the most influential factors influencing osteoporosis. Urinary Co (0.20-6.10 μg/mg creatinine), urinary Tu (0.06-1.93 μg/mg creatinine), blood Cd (0.07-0.50 μg/L), and urinary Hg (0.06-0.75 μg/mg creatinine) levels displayed a distinctive upward trend with risk of osteoporosis as values increased. Our analysis revealed that urinary Co, urinary Tu, blood Cd, and urinary Hg played a significant role in predictability.

Identification of Primary Organophosphate Esters Contributing to Enhanced Risk of Gestational Diabetes Mellitus Based on a Case-Control Study

Epidemiological studies on associations of organophosphate ester (OPE) exposure and gestational diabetes mellitus (GDM) risk, which remain rare and inconclusive, were carried out with a case-control population comprising 287 GDM and 313 non-GDM pregnant women recruited from Tianjin. The GDM group suffered distinctly higher serum concentrations of tri-n-butyl phosphate (TNBP), tri(2-butoxyethyl) phosphate (TBOEP), triphenyl phosphate (TPHP), tri-iso-propyl phosphate (TIPP), and tri(1-chloro-2-propyl) phosphate (TCIPP) than the healthy control group (p < 0.001). Traditional analysis methods employed for either individual or mixture effects found positive correlations (p < 0.05) between the concentrations of five OPEs (i.e., TNBP, TBOEP, TPHP, TIPP, and TCIPP) and the incidence of GDM, while 2-ethylhexyl diphenyl phosphate, tri(1-chloro-2-propyl) phosphate, and bis(2-ethylhexyl) phosphate exhibited opposite effects. Three machine learning methods considering the concurrence of OPE mixture exposure and population characteristics were applied to clarify their relative importance to GDM risk, among which random forest performed the best. Several OPEs, particularly TNBP and TBOEP ranking at the top, made greater contributions than some demographical characteristics, such as prepregnancy body mass index and family history of diabetes, to the occurrence of GDM. This was further validated by another independent case-control population obtained from Hangzhou.

Classification and functional analysis of disulfidptosis-associated genes in sepsis

Sepsis represents a critical condition characterized by multiple-organ dysfunction resulting from inflammatory response to infection. Disulfidptosis is a newly identified type of programmed cell death that is intimately associated with the actin cytoskeleton collapse caused by glucose starvation and disulfide stress, but its role in sepsis is largely unknown. The study was to adopt a diagnostic and prognostic signature for sepsis with disulfidptosis based on the differentially expressed genes (DEGs) between sepsis and healthy people from GEO database. The disulfidptosis hub genes associated with sepsis were identified, and then developed consensus clustering and immune infiltration characteristics. Next, we evaluated disulfidptosis-related risk genes by using LASSO and Random Forest algorithms, and constructed the diagnostic sepsis model by nomogram. Finally, immune infiltration, GSVA analysis and mRNA-miRNA networks based on disulfidptosis-related DEGs were screened. There are five upregulated disulfidptosis-related genes and seven downregulated genes were filtered out. The six intersection disulfidptosis-related genes including LRPPRC, SLC7A11, GLUT, MYH9, NUBPL and GYS1 exhibited higher predictive ability for sepsis with an accuracy of 99.7%. In addition, the expression patterns of the critical genes were validated. The study provided a comprehensive view of disulfidptosis-based signatures to predict the prognosis, biological features and potential treatment directions for sepsis.

Effects of environmental phenols on eGFR: machine learning modeling methods applied to cross-sectional studies

Purpose

Limited investigation is available on the correlation between environmental phenols' exposure and estimated glomerular filtration rate (eGFR). Our target is established a robust and explainable machine learning (ML) model that associates environmental phenols' exposure with eGFR.

Methods

Our datasets for constructing the associations between environmental phenols' and eGFR were collected from the National Health and Nutrition Examination Survey (NHANES, 2013-2016). Five ML models were contained and fine-tuned to eGFR regression by phenols' exposure. Regression evaluation metrics were used to extract the limitation of the models. The most effective model was then utilized for regression, with interpretation of its features carried out using shapley additive explanations (SHAP) and the game theory python package to represent the model's regression capacity.

Results

The study identified the top-performing random forest (RF) regressor with a mean absolute error of 0.621 and a coefficient of determination of 0.998 among 3,371 participants. Six environmental phenols with eGFR in linear regression models revealed that the concentrations of triclosan (TCS) and bisphenol S (BPS) in urine were positively correlated with eGFR, and the correlation coefficients were β = 0.010 (p = 0.026) and β = 0.007 (p = 0.004) respectively. SHAP values indicate that BPS (1.38), bisphenol F (BPF) (0.97), 2,5-dichlorophenol (0.87), TCS (0.78), BP3 (0.60), bisphenol A (BPA) (0.59) and 2,4-dichlorophenol (0.47) in urinary contributed to the model.

Conclusion

The RF model was efficient in identifying a correlation between phenols' exposure and eGFR among United States NHANES 2013-2016 participants. The findings indicate that BPA, BPF, and BPS are inversely associated with eGFR.

Association of serum metal levels with type 2 diabetes: A prospective cohort and mediating effects of metabolites analysis in Chinese population

Several studies have suggested an association between exposure to various metals and the onset of type 2 diabetes (T2D). However, the results vary across different studies. We aimed to investigate the associations between serum metal concentrations and the risk of developing T2D among 8734 participants using a prospective cohort study design. We utilized inductively coupled plasmamass spectrometry (ICP-MS) to assess the serum concentrations of 27 metals. Cox regression was applied to calculate the hazard ratios (HRs) for the associations between serum metal concentrations on the risk of developing T2D. Additionally, 196 incident T2D cases and 208 healthy control participants were randomly selected for serum metabolite measurement using an untargeted metabolomics approach to evaluate the mediating role of serum metabolite in the relationship between serum metal concentrations and the risk of developing T2D with a nested casecontrol study design. In the cohort study, after Bonferroni correction, the serum concentrations of zinc (Zn), mercury (Hg), and thallium (Tl) were positively associated with the risk of developing T2D, whereas the serum concentrations of manganese (Mn), molybdenum (Mo), barium (Ba), lutetium (Lu), and lead (Pb) were negatively associated with the risk of developing T2D. After adding these eight metals, the predictive ability increased significantly compared with that of the traditional clinical model (AUC: 0.791 vs. 0.772, P=8.85×10-5). In the nested casecontrol study, a machine learning analysis revealed that the serum concentrations of 14 out of 1579 detected metabolites were associated with the risk of developing T2D. According to generalized linear regression models, 7 of these metabolites were significantly associated with the serum concentrations of the identified metals. The mediation analysis showed that two metabolites (2-methyl-1,2-dihydrophthalazin-1-one and mestranol) mediated 46.81% and 58.70%, respectively, of the association between the serum Pb concentration and the risk of developing T2D. Our study suggested that serum Mn, Zn, Mo, Ba, Lu, Hg, Tl, and Pb were associated with T2D risk. Two metabolites mediated the associations between the serum Pb concentration and the risk of developing T2D.

A machine learning model predicts stroke associated with blood cadmium level

Stroke is the leading cause of death and disability worldwide. Cadmium is a prevalent environmental toxicant that may contribute to cardiovascular disease, including stroke. We aimed to build an effective and interpretable machine learning (ML) model that links blood cadmium to the identification of stroke. Our data exploring the association between blood cadmium and stroke came from the National Health and Nutrition Examination Survey (NHANES, 2013-2014). In total, 2664 participants were eligible for this study. We divided these data into a training set (80%) and a test set (20%). To analyze the relationship between blood cadmium and stroke, a multivariate logistic regression analysis was performed. We constructed and tested five ML algorithms including K-nearest neighbor (KNN), decision tree (DT), logistic regression (LR), multilayer perceptron (MLP), and random forest (RF). The best-performing model was selected to identify stroke in US adults. Finally, the features were interpreted using the Shapley Additive exPlanations (SHAP) tool. In the total population, participants in the second, third, and fourth quartiles had an odds ratio of 1.32 (95% CI 0.55, 3.14), 1.65 (95% CI 0.71, 3.83), and 2.67 (95% CI 1.10, 6.49) for stroke compared with the lowest reference group for blood cadmium, respectively. This blood cadmium-based LR approach demonstrated the greatest performance in identifying stroke (area under the operator curve: 0.800, accuracy: 0.966). Employing interpretable methods, we found blood cadmium to be a notable contributor to the predictive model. We found that blood cadmium was positively correlated with stroke risk and that stroke risk from cadmium exposure could be effectively predicted by using ML modeling.

Greenness, air pollution, and temperature exposure effects in predicting premature mortality and morbidity: A small-area study using spatial random forest model

BACKGROUND

Although studies have provided negative impacts of air pollution, heat or cold exposure on mortality and morbidity, and positive effects of increased greenness on reducing them, a few studies have focused on exploring combined and synergetic effects of these exposures in predicting these health outcomes, and most had ignored the spatial autocorrelation in analyzing their health effects. This study aims to investigate the health effects of air pollution, greenness, and temperature exposure on premature mortality and morbidity within a spatial machine-learning modeling framework.

METHODS

Years of potential life lost reflecting premature mortality and comparative illness and disability ratio reflecting chronic morbidity from 1673 small areas covering Greater Manchester for the year 2008-2013 obtained. Average annual levels of NO2 concentration, normalized difference vegetation index (NDVI) representing greenness, and annual average air temperature were utilized to assess exposure in each area. These exposures were linked to health outcomes using non-spatial and spatial random forest (RF) models while accounting for spatial autocorrelation.

RESULTS

Spatial-RF models provided the best predictive accuracy when accounted for spatial autocorrelation. Among the exposures considered, air pollution emerged as the most influential in predicting mortality and morbidity, followed by NDVI and temperature exposure. Nonlinear exposure-response relations were observed, and interactions between exposures illustrated specific ranges or sweet and sour spots of exposure thresholds where combined effects either exacerbate or moderate health conditions.

CONCLUSION

Air pollution exposure had a greater negative impact on health compared to greenness and temperature exposure. Combined exposure effects may indicate the highest influence of premature mortality and morbidity burden.

Exploring the relationship between heavy metals and diabetic retinopathy: a machine learning modeling approach

Diabetic retinopathy (DR) is one of the leading causes of adult blindness in the United States. Although studies applying traditional statistical methods have revealed that heavy metals may be essential environmental risk factors for diabetic retinopathy, there is a lack of analyses based on machine learning (ML) methods to adequately explain the complex relationship between heavy metals and DR and the interactions between variables. Based on characteristic variables of participants with and without DR and heavy metal exposure data obtained from the NHANES database (2003-2010), a ML model was developed for effective prediction of DR. The best predictive model for DR was selected from 11 models by receiver operating characteristic curve (ROC) analysis. Further permutation feature importance (PFI) analysis, partial dependence plots (PDP) analysis, and SHapley Additive exPlanations (SHAP) analysis were used to assess the model capability and key influencing factors. A total of 1042 eligible individuals were randomly assigned to two groups for training and testing set of the prediction model. ROC analysis showed that the k-nearest neighbour (KNN) model had the highest prediction performance, achieving close to 100% accuracy in the testing set. Urinary Sb level was identified as the critical heavy metal affecting the predicted risk of DR, with a contribution weight of 1.730632 ± 1.791722, which was much higher than that of other heavy metals and baseline variables. The results of the PDP analysis and the SHAP analysis also indicated that antimony (Sb) had a more significant effect on DR. The interaction between age and Sb was more significant compared to other variables and metal pairs. We found that Sb could serve as a potential predictor of DR and that Sb may influence the development of DR by mediating cellular and systemic senescence. The study revealed that monitoring urinary Sb levels can be useful for early non-invasive screening and intervention in DR development, and also highlighted the important role of constructed ML models in explaining the effects of heavy metal exposure on DR.

Risk of papillary thyroid carcinoma and nodular goiter associated with exposure to semi-volatile organic compounds: A multi-pollutant assessment based on machine learning algorithms

BACKGROUND

Exposure to semi-volatile organic compounds (SVOCs) may link to thyroid nodule risk, but studies of mixed-SVOCs exposure effects are lacking. Traditional analytical methods are inadequate for dealing with mixed exposures, while machine learning (ML) seems to be a good way to fill the gaps in the field of environmental epidemiology research.

OBJECTIVES

Different ML algorithms were used to explore the relationship between mixed-SVOCs exposure and thyroid nodule.

METHODS

A 1:1:1 age- and gender-matched case-control study was conducted in which 96 serum SVOCs were measured in 50 papillary thyroid carcinoma (PTC), 50 nodular goiters (NG), and 50 controls. Different ML techniques such as Random Forest, AdaBoost were selected based on their predictive power, and variables were selected based on their weights in the models. Weighted quantile sum (WQS) regression and Bayesian kernel machine regression (BKMR) were used to assess the mixed effects of the SVOCs exposure on thyroid nodule.

RESULTS

Forty-three of 96 SVOCs with detection rate >80 % were included in the analysis. ML algorithms showed a consistent selection of SVOCs associated with thyroid nodule. Fluazifop-butyl and fenpropathrin are positively associated with PTC and NG in single compound models (all P < 0.05). WQS model shows that exposure to mixed-SVOCs was associated with an increased risk of PTC and NG, with the mixture dominated by fenpropathrin, followed by fluazifop-butyl and propham. In the BKMR model, mixtures showed a significant positive association with thyroid nodule risk at high exposure levels, and fluazifop-butyl showed positive effects associated with PTC and NG.

CONCLUSION

This study confirms the feasibility of ML methods for variable selection in high-dimensional complex data and showed that mixed exposure to SVOCs was associated with increased risk of PTC and NG. The observed association was primarily driven by fluazifop-butyl and fenpropathrin. The findings warranted further investigation.

Identification of influence factors in overweight population through an interpretable risk model based on machine learning: a large retrospective cohort

BACKGROUND

The identification of associated overweight risk factors is crucial to future health risk predictions and behavioral interventions. Several consensus problems remain in machine learning, such as cross-validation, and the resulting model may suffer from overfitting or poor interpretability.

METHODS

This study employed nine commonly used machine learning methods to construct overweight risk models. The general community are the target of this study, and a total of 10,905 Chinese subjects from Ningde City in Fujian province, southeast China, participated. The best model was selected through appropriate verification and validation and was suitably explained.

RESULTS

The overweight risk models employing machine learning exhibited good performance. It was concluded that CatBoost, which is used in the construction of clinical risk models, may surpass previous machine learning methods. The visual display of the Shapley additive explanation value for the machine model variables accurately represented the influence of each variable in the model.

CONCLUSIONS

The construction of an overweight risk model using machine learning may currently be the best approach. Moreover, CatBoost may be the best machine learning method. Furthermore, combining Shapley's additive explanation and machine learning methods can be effective in identifying disease risk factors for prevention and control.

An artificial neural network model for evaluating the risk of hyperuricaemia in type 2 diabetes mellitus

Type 2 diabetes with hyperuricaemia may lead to gout, kidney damage, hypertension, coronary heart disease, etc., further aggravating the condition of diabetes as well as adding to the medical and financial burden. To construct a risk model for hyperuricaemia in patients with type 2 diabetes mellitus based on artificial neural network, and to evaluate the effectiveness of the risk model to provide directions for the prevention and control of the disease in this population. From June to December 2022, 8243 patients with type 2 diabetes were recruited from six community service centers for questionnaire and physical examination. Secondly, the collected data were used to select suitable variables and based on the comparison results, logistic regression was used to screen the variable characteristics. Finally, three risk models for evaluating the risk of hyperuricaemia in type 2 diabetes mellitus were developed using an artificial neural network algorithm and evaluated for performance. A total of eleven factors affecting the development of hyperuricaemia in patients with type 2 diabetes mellitus in this study, including gender, waist circumference, diabetes medication use, diastolic blood pressure, γ-glutamyl transferase, blood urea nitrogen, triglycerides, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, fasting glucose and estimated glomerular filtration rate. Among the generated models, baseline & biochemical risk model had the best performance with cutoff, area under the curve, accuracy, recall, specificity, positive likelihood ratio, negative likelihood ratio, precision, negative predictive value, KAPPA and F1-score were 0.488, 0.744, 0.689, 0.625, 0.749, 2.489, 0.501, 0.697, 0.684, 0.375 and 0.659. In addition, its Brier score was 0.169 and the calibration curve also showed good agreement between fitting and observation. The constructed artificial neural network model has better efficacy and facilitates the reduction of the harm caused by type 2 diabetes mellitus combined with hyperuricaemia.

Toward reliable diabetes prediction: Innovations in data engineering and machine learning applications

OBJECTIVE

Diabetes is a metabolic disorder that causes the risk of stroke, heart disease, kidney failure, and other long-term complications because diabetes generates excess sugar in the blood. Machine learning (ML) models can aid in diagnosing diabetes at the primary stage. So, we need an efficient ML model to diagnose diabetes accurately.

METHODS

In this paper, an effective data preprocessing pipeline has been implemented to process the data and random oversampling to balance the data, handling the imbalance distributions of the observational data more sophisticatedly. We used four different diabetes datasets to conduct our experiments. Several ML algorithms were used to determine the best models to predict diabetes faultlessly.

RESULTS

The performance analysis demonstrates that among all ML algorithms, random forest surpasses the current works with an accuracy rate of 86% and 98.48% for Dataset 1 and Dataset 2; extreme gradient boosting and decision tree surpass with an accuracy rate of 99.27% and 100% for Dataset 3 and Dataset 4, respectively. Our proposal can increase accuracy by 12.15% compared to the model without preprocessing.

CONCLUSIONS

This excellent research finding indicates that the proposed models might be employed to produce more accurate diabetes predictions to supplement current preventative interventions to reduce the incidence of diabetes and its associated costs.

Comprehensive analysis of necroptosis-related genes in renal ischemia-reperfusion injury

Background

Oxidative stress is the primary cause of ischemia-reperfusion injury (IRI) in kidney transplantation, leading to delayed graft function (DGF) and implications on patient health. Necroptosis is believed to play a role in renal IRI. This research presents a comprehensive analysis of necroptosis-related genes and their functional implications in the context of IRI in renal transplantation.

Methods

The necroptosis-related differentially expressed genes (NR-DEGs) were identified using gene expression data from pre- and post-reperfusion renal biopsies, and consensus clustering analysis was performed to distinguish necroptosis-related clusters. A predictive model for DGF was developed based on the NR-DEGs and patients were divided into high- and low-risk groups. We investigated the differences in functional enrichment and immune infiltration between different clusters and risk groups and further validated them in single-cell RNA-sequencing (scRNA-seq) data. Finally, we verified the expression changes of NR-DEGs in an IRI mouse model.

Results

Five NR-DEGs were identified and were involved in various biological processes. The renal samples were further stratified into two necroptosis-related clusters (C1 and C2) showing different occurrences of DGF. The predictive model had a reliable performance in identifying patients at higher risk of DGF with the area under the curve as 0.798. Additionally, immune infiltration analysis indicated more abundant proinflammatory cells in the high-risk group, which was also found in C2 cluster with more DGF patients. Validation of NR-DEG in scRNA-seq data further supported their involvement in immune cells. Lastly, the mouse model validated the up-regulation of NR-DEGs after IR and indicated the correlations with kidney function markers.

Conclusions

Our research provides valuable insights into the identification and functional characterization of NR-DEGs in the context of renal transplantation and sheds light on their involvement in immune responses and the progression of IRI and DGF.

Association between co-exposure to phenols, phthalates, and polycyclic aromatic hydrocarbons with the risk of frailty

The phenomenon of population aging has brought forth the challenge of frailty. Nevertheless, the contribution of environmental exposure to frailty remains ambiguous. Our objective was to investigate the association between phenols, phthalates (PAEs), and polycyclic aromatic hydrocarbons (PAHs) with frailty. We constructed a 48-item frailty index using data from the National Health and Nutrition Examination Survey (NHANES). The exposure levels of 20 organic contaminants were obtained from the survey circle between 2005 and 2016. The association between individual organic contaminants and the frailty index was assessed using negative binomial regression models. The combined effect of organic contaminants was examined using weighted quantile sum (WQS) regression. Dose-response patterns were modeled using generalized additive models (GAMs). Additionally, an interpretable machine learning approach was employed to develop a predictive model for the frailty index. A total of 1566 participants were included in the analysis. Positive associations were observed between exposure to MIB, P02, ECP, MBP, MHH, MOH, MZP, MC1, and P01 with the frailty index. WQS regression analysis revealed a significant increase in the frailty index with higher levels of the mixture of organic contaminants (aOR, 1.12; 95% CI, 1.05-1.20; p < 0.001), with MIB, ECP, COP, MBP, P02, and P01 identified as the major contributors. Dose-response relationships were observed between MIB, ECP, MBP, P02, and P01 exposure with an increased risk of frailty (both with p < 0.05). The developed predictive model based on organic contaminants exposure demonstrated high performance, with an R2 of 0.9634 and 0.9611 in the training and testing sets, respectively. Furthermore, the predictive model suggested potential synergistic effects in the MIB-MBP and P01-P02 pairs. Taken together, these findings suggest a significant association between exposure to phthalates and PAHs with an increased susceptibility to frailty.

Identification of dietary components in association with abdominal aortic calcification

The precise impact of dietary components on vascular health remains incompletely understood. To identify the dietary components and their associations with abdominal aortic calcification (AAC), the data from NHANES was employed in this cross-sectional study. The LASSO method and logistic regression were utilized to identify dietary components that exhibited the strongest association with AAC. Grouped WQS regression analysis was employed to evaluate the combined effects of dietary components on AAC. Furthermore, principal component analysis was employed to identify the primary dietary patterns in the study population. The present analysis included 1862 participants, from whom information on 35 dietary macro- and micronutrient components was obtained through 24-hour dietary recall interviews. The assessment of AAC was performed utilizing dual-energy X-ray absorptiometry. The LASSO method identified 10 dietary components that were associated with AAC. Total protein, total fiber, vitamin A, and β-cryptoxanthin exhibited a negative association with AAC. Compared to the first quartile, the adjusted odds ratios (95% CIs) for the highest quartile were 0.59 (0.38, 0.93), 0.63 (0.42, 0.93), 0.59 (0.41, 0.84), and 0.68 (0.48, 0.94), respectively. Grouped WQS regression demonstrated a positive association between the lipid group and AAC (aOR: 1.29; 95% CI: 1.12, 1.50), while the proteins and phytochemical group exhibited a negative association with AAC (aOR: 0.69; 95% CI: 0.58, 0.82). For the dietary pattern analysis, high adherence to the plant-based pattern (aOR: 0.62; 95% CI: 0.44, 0.88) was associated with a lower risk of AAC, whereas the caffeine and theobromine pattern (aOR: 1.73; 95% CI: 1.25, 2.41) was associated with a higher risk of AAC. The findings of this study indicate that adopting a dietary pattern characterized by high levels of protein and plant-based foods, as well as reduced levels of fat, may offers potential advantages for the prevention of AAC.

Application of Machine Learning Models for Early Detection and Accurate Classification of Type 2 Diabetes

Early detection of diabetes is essential to prevent serious complications in patients. The purpose of this work is to detect and classify type 2 diabetes in patients using machine learning (ML) models, and to select the most optimal model to predict the risk of diabetes. In this paper, five ML models, including K-nearest neighbor (K-NN), Bernoulli Naïve Bayes (BNB), decision tree (DT), logistic regression (LR), and support vector machine (SVM), are investigated to predict diabetic patients. A Kaggle-hosted Pima Indian dataset containing 768 patients with and without diabetes was used, including variables such as number of pregnancies the patient has had, blood glucose concentration, diastolic blood pressure, skinfold thickness, body insulin levels, body mass index (BMI), genetic background, diabetes in the family tree, age, and outcome (with/without diabetes). The results show that the K-NN and BNB models outperform the other models. The K-NN model obtained the best accuracy in detecting diabetes, with 79.6% accuracy, while the BNB model obtained 77.2% accuracy in detecting diabetes. Finally, it can be stated that the use of ML models for the early detection of diabetes is very promising.

Effects of heavy metal exposure on hypertension: A machine learning modeling approach

Heavy metal exposure is a common risk factor for hypertension. To develop an interpretable predictive machine learning (ML) model for hypertension based on levels of heavy metal exposure, data from the NHANES (2003-2016) were employed. Random forest (RF), support vector machine (SVM), decision tree (DT), multilayer perceptron (MLP), ridge regression (RR), AdaBoost (AB), gradient boosting decision tree (GBDT), voting classifier (VC), and K-nearest neighbour (KNN) algorithms were utilized to generate an optimal predictive model for hypertension. Three interpretable methods, the permutation feature importance analysis, partial dependence plot (PDP), and Shapley additive explanations (SHAP) methods, were integrated into a pipeline and embedded in ML for model interpretation. A total of 9005 eligible individuals were randomly allocated into two distinct sets for predictive model training and validation. The results showed that among the predictive models, the RF model demonstrated the highest performance, achieving an accuracy rate of 77.40% in the validation set. The AUC and F1 score for the model were 0.84 and 0.76, respectively. Blood Pb, urinary Cd, urinary Tl, and urinary Co levels were identified as the main influencers of hypertension, and their contribution weights were 0.0504 ± 0.0482, 0.0389 ± 0.0256, 0.0307 ± 0.0179, and 0.0296 ± 0.0162, respectively. Blood Pb (0.55-2.93 μg/dL) and urinary Cd (0.06-0.15 μg/L) levels exhibited the most pronounced upwards trend with the risk of hypertension within a specific value range, while urinary Tl (0.06-0.26 μg/L) and urinary Co (0.02-0.32 μg/L) levels demonstrated a declining trend with hypertension. The findings on the synergistic effects indicated that Pb and Cd were the primary determinants of hypertension. Our findings underscore the predictive value of heavy metals for hypertension. By utilizing interpretable methods, we discerned that Pb, Cd, Tl, and Co emerged as noteworthy contributors within the predictive model.

A machine learning-based diagnosis modelling of type 2 diabetes mellitus with environmental metal exposure

BACKGROUND AND OBJECTIVE

Increasing and compelling evidence has been proved that urinary and dietary metal exposure are underappreciated but potentially modifiable biomarkers for type 2 diabetes mellitus (T2DM). The aims of this study were (1) to identify the key potential biomarkers which contributed to T2DM with effective and parsimonious features and (2) to assess the utility of baseline variables and metal exposure in the diagnosis of T2DM.

METHODS

Based on the National Health and Nutrition Examination Survey (NHANES), we selected 9822 screening records with 82 significant variables covering demographics, lifestyle, anthropometric measures, diet and metal exposure for this study. Combining extreme gradient boosting (XGBoost), random forest and light gradient boosting machine (lightGBM), a soft voting ensemble model was proposed to measure the importance of 82 features. With this soft voting ensemble model and variance inflation factor (VIF), strong multicollinear features with low importance scores were further removed from candidate biomarkers. Then, a soft voting ensemble classifier was adopted to demonstrate the efficiency of the proposed feature selection method.

RESULTS

With the novel feature selection method, 12 baseline variables and 3 metal variables were selected to detect patients at risk for T2DM in our study. For metal variables, the dietary copper (Cu), urinary cadmium (Cd) and urinary mercury (Hg) metals were selected as the most remarkable metal exposure and the corresponding P-values were all less than 0.05. In a classification model of T2DM with 12 baseline biomarkers, the addition of 3 metal exposure improved the classification accuracy of T2DM from a traditional area under the curve (AUC) 0.792 of the receiver operating characteristic (ROC) to an AUC 0.847.

CONCLUSIONS

This was the first demonstration of T2DM classification with machine learning under urinary and dietary metal exposure. Improved prediction precision illustrated the effectiveness of the proposed machine learning-based diagnosis model facilitated lifestyle/dietary intervention for T2DM prevention.

Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review

Diabetes as a metabolic illness can be characterized by increased amounts of blood glucose. This abnormal increase can lead to critical detriment to the other organs such as the kidneys, eyes, heart, nerves, and blood vessels. Therefore, its prediction, prognosis, and management are essential to prevent harmful effects and also recommend more useful treatments. For these goals, machine learning algorithms have found considerable attention and have been developed successfully. This review surveys the recently proposed machine learning (ML) and deep learning (DL) models for the objectives mentioned earlier. The reported results disclose that the ML and DL algorithms are promising approaches for controlling blood glucose and diabetes. However, they should be improved and employed in large datasets to affirm their applicability.

Bioactivity Guided Study for the Isolation and Identification of Antidiabetic Compounds from Edible Seaweed-Ulva reticulata

Managing diabetes is challenging due to the complex physiology of the disease and the numerous complications associated with it. As part of the ongoing search for antidiabetic chemicals, marine algae have been demonstrated to be an excellent source due to their medicinal properties. In this study, Ulva reticulata extracts were investigated for their anti-diabetic effect by examining its inhibitory effects on α-amylase, α-glucosidase, and DPP-IV and antioxidant (DPPH) potential in vitro and its purified fraction using animal models. Among the various solvents used, the Methanolic extract of Ulva reticulata (MEUR) displayed the highest antidiabetic activity in both in vitro and in vivo; it showed no cytotoxicity and hence was subjected to bioassay-guided chromatographic separation. Among the seven isolated fractions (F1 to F7), the F4 (chloroform) fraction exhibited substantial total phenolic content (65.19 μg mL^-1) and total flavonoid content (20.33 μg mL^-1), which showed the promising inhibition against α-amylase (71.67%) and α-glucosidase (38.01%). Active fraction (F4) was further purified using column chromatography, subjected to thin-layer chromatography (TLC), and characterized by spectroscopy techniques. Upon structural elucidation, five distinct compounds, namely, Nonane, Hexadecanoic acid, 1-dodecanol, Cyclodecane methyl, and phenol, phenol, 3,5-bis(1,1-dimethylethyl) were identified. The antidiabetic mechanism of active fraction (F4) was further investigated using various in vitro and in vivo models. The results displayed that in in vitro both 1 and 24 h in vitro cultures, the active fraction (F4) at a concentration of 100 μg mL^-1 demonstrated maximum glucose-induced insulin secretion at 4 mM (0.357 and 0.582 μg mL^-1) and 20 mM (0.848 and 1.032 μg mL^-1). The active fraction (F4) reduces blood glucose levels in normoglycaemic animals and produces effects similar to that of standard acarbose. Active fraction (F4) also demonstrated outstanding hypoglycaemic activity in hyperglycemic animals at a dose of 10 mg/kg B.wt. In the STZ-induced diabetic rat model, the active fraction (F4) showed a (61%) reduction in blood glucose level when compared to the standard drug glibenclamide (68%). The results indicate that the marine algae Ulva reticulata is a promising candidate for managing diabetes by inhibiting carbohydrate metabolizing enzymes and promoting insulin secretion.

Metabolomic Selection in the Progression of Type 2 Diabetes Mellitus: A Genetic Algorithm Approach

According to the World Health Organization (WHO), type 2 diabetes mellitus (T2DM) is a result of the inefficient use of insulin by the body. More than 95% of people with diabetes have T2DM, which is largely due to excess weight and physical inactivity. This study proposes an intelligent feature selection of metabolites related to different stages of diabetes, with the use of genetic algorithms (GA) and the implementation of support vector machines (SVMs), K-Nearest Neighbors (KNNs) and Nearest Centroid (NEARCENT) and with a dataset obtained from the Instituto Mexicano del Seguro Social with the protocol name of the following: "Análisis metabolómico y transcriptómico diferencial en orina y suero de pacientes pre diabéticos, diabéticos y con nefropatía diabética para identificar potenciales biomarcadores pronósticos de daño renal" (differential metabolomic and transcriptomic analyses in the urine and serum of pre-diabetic, diabetic and diabetic nephropathy patients to identify potential prognostic biomarkers of kidney damage). In order to analyze which machine learning (ML) model is the most optimal for classifying patients with some stage of T2DM, the novelty of this work is to provide a genetic algorithm approach that detects significant metabolites in each stage of progression. More than 100 metabolites were identified as significant between all stages; with the data analyzed, the average accuracies obtained in each of the five most-accurate implementations of genetic algorithms were in the range of 0.8214-0.9893 with respect to average accuracy, providing a precise tool to use in detections and backing up a diagnosis constructed entirely with metabolomics. By providing five potential biomarkers for progression, these extremely significant metabolites are as follows: "Cer(d18:1/24:1) i2", "PC(20:3-OH/P-18:1)", "Ganoderic acid C2", "TG(16:0/17:1/18:1)" and "GPEtn(18:0/20:4)".

Association between exposure to organophosphorus pesticides and the risk of diabetes among US Adults: Cross-sectional findings from the National Health and Nutrition Examination Survey

OBJECTIVE

Organophosphorus pesticides (OPPs) are commonly used pesticides across the world, however there is little epidemiological evidence linking their exposure to diabetes. Hence, this study aimed at investigating the effect of OPP exposure on the prevalence of diabetes in American adults.

METHODS

Adults (≥20 years old) were eligible for this study from the National Health and Nutrition Examination Survey (NHANES). Multivariate logistic regression model was employed to explore the associations of six main urinary OPPs metabolites with diabetes. Subgroup analyses were performed by age and gender. Combined effect of OPPs metabolites on the overall association with diabetes was evaluated by weighted quantile sum regression (WQS). Furthermore, Bayesian kernel machine regression (BKMR) model was implemented to explore joint effect of multiple OPPs metabolites on diabetes.

RESULTS

Ultimately, 6,593 adults were included in our analysis. Of them, 1,044 participants were determined as diabetes patients. The results of logistic regression shown that urinary OPPs metabolites concentrations, whether taken as continuous variables or quantiles, were in positive correlation with diabetes. Notably, the p for trend of diethylphosphate (DEP), a kind of OPPs metabolites, was less than 0.05 indicated that a linear trend may exist between levels of DEP and prevalence of diabetes among adults while this trend was not obversed in other OPPs metabolites. In the WQS model, combined exposure of OPPs metabolites had a significantly positive association with diabetes (OR: 1.057; 95% CI: 1.002, 1.114) and diethylphosphate (36.84%) made the largest contributor to the WQS index. The result of BKMR also suggested a positive trend of association between mixed OPPs metabolites and diabetes.

CONCLUSION

Our results add credibility to the argument that OPP exposure might trigger diabetes. Certainly, prospective data are required to corroborate our findings.

Predicting the Risk of Incident Type 2 Diabetes Mellitus in Chinese Elderly Using Machine Learning Techniques

Early identification of individuals at high risk of diabetes is crucial for implementing early intervention strategies. However, algorithms specific to elderly Chinese adults are lacking. The aim of this study is to build effective prediction models based on machine learning (ML) for the risk of type 2 diabetes mellitus (T2DM) in Chinese elderly. A retrospective cohort study was conducted using the health screening data of adults older than 65 years in Wuhan, China from 2018 to 2020. With a strict data filtration, 127,031 records from the eligible participants were utilized. Overall, 8298 participants were diagnosed with incident T2DM during the 2-year follow-up (2019–2020). The dataset was randomly split into training set (n = 101,625) and test set (n = 25,406). We developed prediction models based on four ML algorithms: logistic regression (LR), decision tree (DT), random forest (RF), and extreme gradient boosting (XGBoost). Using LASSO regression, 21 prediction features were selected. The Random under-sampling (RUS) was applied to address the class imbalance, and the Shapley Additive Explanations (SHAP) was used to calculate and visualize feature importance. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy. The XGBoost model achieved the best performance (AUC = 0.7805, sensitivity = 0.6452, specificity = 0.7577, accuracy = 0.7503). Fasting plasma glucose (FPG), education, exercise, gender, and waist circumference (WC) were the top five important predictors. This study showed that XGBoost model can be applied to screen individuals at high risk of T2DM in the early phrase, which has the strong potential for intelligent prevention and control of diabetes. The key features could also be useful for developing targeted diabetes prevention interventions.

An Interactive Online App for Predicting Diabetes via Machine Learning from Environment-Polluting Chemical Exposure Data

The early prediction and identification of risk factors for diabetes may prevent or delay diabetes progression. In this study, we developed an interactive online application that provides the predictive probabilities of prediabetes and diabetes in 4 years based on a Bayesian network (BN) classifier, which is an interpretable machine learning technique. The BN was trained using a dataset from the Ansung cohort of the Korean Genome and Epidemiological Study (KoGES) in 2008, with a follow-up in 2012. The dataset contained not only traditional risk factors (current diabetes status, sex, age, etc.) for future diabetes, but it also contained serum biomarkers, which quantified the individual level of exposure to environment-polluting chemicals (EPC). Based on accuracy and the area under the curve (AUC), a tree-augmented BN with 11 variables derived from feature selection was used as our prediction model. The online application that implemented our BN prediction system provided a tool that performs customized diabetes prediction and allows users to simulate the effects of controlling risk factors for the future development of diabetes. The prediction results of our method demonstrated that the EPC biomarkers had interactive effects on diabetes progression and that the use of the EPC biomarkers contributed to a substantial improvement in prediction performance.

Development and validation of a machine learning-augmented algorithm for diabetes screening in community and primary care settings: A population-based study

BACKGROUND

Opportunely screening for diabetes is crucial to reduce its related morbidity, mortality, and socioeconomic burden. Machine learning (ML) has excellent capability to maximize predictive accuracy. We aim to develop ML-augmented models for diabetes screening in community and primary care settings.

METHODS

8425 participants were involved from a population-based study in Hubei, China since 2011. The dataset was split into a development set and a testing set. Seven different ML algorithms were compared to generate predictive models. Non-laboratory features were employed in the ML model for community settings, and laboratory test features were further introduced in the ML+lab models for primary care. The area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (auPR), and the average detection costs per participant of these models were compared with their counterparts based on the New China Diabetes Risk Score (NCDRS) currently recommended for diabetes screening.

RESULTS

The AUC and auPR of the ML model were 0·697and 0·303 in the testing set, seemingly outperforming those of NCDRS by 10·99% and 64·67%, respectively. The average detection cost of the ML model was 12·81% lower than that of NCDRS with the same sensitivity (0·72). Moreover, the average detection cost of the ML+FPG model is the lowest among the ML+lab models and less than that of the ML model and NCDRS+FPG model.

CONCLUSION

The ML model and the ML+FPG model achieved higher predictive accuracy and lower detection costs than their counterpart based on NCDRS. Thus, the ML-augmented algorithm is potential to be employed for diabetes screening in community and primary care settings.

Machine learning and deep learning predictive models for type 2 diabetes: a systematic review

Diabetes Mellitus is a severe, chronic disease that occurs when blood glucose levels rise above certain limits. Over the last years, machine and deep learning techniques have been used to predict diabetes and its complications. However, researchers and developers still face two main challenges when building type 2 diabetes predictive models. First, there is considerable heterogeneity in previous studies regarding techniques used, making it challenging to identify the optimal one. Second, there is a lack of transparency about the features used in the models, which reduces their interpretability. This systematic review aimed at providing answers to the above challenges. The review followed the PRISMA methodology primarily, enriched with the one proposed by Keele and Durham Universities. Ninety studies were included, and the type of model, complementary techniques, dataset, and performance parameters reported were extracted. Eighteen different types of models were compared, with tree-based algorithms showing top performances. Deep Neural Networks proved suboptimal, despite their ability to deal with big and dirty data. Balancing data and feature selection techniques proved helpful to increase the model's efficiency. Models trained on tidy datasets achieved almost perfect models.