Development and Validation of an Insulin Resistance Predicting Model Using a Machine-Learning Approach in a Population-Based Cohort in Korea

Application of machine learning in affordable and accessible insulin management for type 1 and 2 diabetes: A comprehensive review

Proper insulin management is vital for maintaining stable blood sugar levels and preventing complications associated with diabetes. However, the soaring costs of insulin present significant challenges to ensuring affordable management. This paper conducts a comprehensive review of current literature on the application of machine learning (ML) in insulin management for diabetes patients, particularly focusing on enhancing affordability and accessibility within the United States. The review encompasses various facets of insulin management, including dosage calculation and response, prediction of blood glucose and insulin sensitivity, initial insulin estimation, resistance prediction, treatment adherence, complications, hypoglycemia prediction, and lifestyle modifications. Additionally, the study identifies key limitations in the utilization of ML within the insulin management literature and suggests future research directions aimed at furthering accessible and affordable insulin treatments. These proposed directions include exploring insurance coverage, optimizing insulin type selection, assessing the impact of biosimilar insulin and market competition, considering mental health factors, evaluating insulin delivery options, addressing cost-related issues affecting insulin usage and adherence, and selecting appropriate patient cost-sharing programs. By examining the potential of ML in addressing insulin management affordability and accessibility, this work aims to envision improved and cost-effective insulin management practices. It not only highlights existing research gaps but also offers insights into future directions, guiding the development of innovative solutions that have the potential to revolutionize insulin management and benefit patients reliant on this life-saving treatment.

Effects of Various Heavy Metal Exposures on Insulin Resistance in Non-diabetic Populations: Interpretability Analysis from Machine Learning Modeling Perspective

Increasing and compelling evidence has been proved that heavy metal exposure is involved in the development of insulin resistance (IR). We trained an interpretable predictive machine learning (ML) model for IR in the non-diabetic populations based on levels of heavy metal exposure. A total of 4354 participants from the NHANES (2003-2020) with complete information were randomly divided into a training set and a test set. Twelve ML algorithms, including random forest (RF), XGBoost (XGB), logistic regression (LR), GaussianNB (GNB), ridge regression (RR), support vector machine (SVM), multilayer perceptron (MLP), decision tree (DT), AdaBoost (AB), Gradient Boosting Decision Tree (GBDT), Voting Classifier (VC), and K-Nearest Neighbour (KNN), were constructed for IR prediction using the training set. Among these models, the RF algorithm had the best predictive performance, showing an accuracy of 80.14%, an AUC of 0.856, and an F1 score of 0.74 in the test set. We embedded three interpretable methods, the permutation feature importance analysis, partial dependence plot (PDP), and Shapley additive explanations (SHAP) in RF model for model interpretation. Urinary Ba, urinary Mo, blood Pb, and blood Cd levels were identified as the main influencers of IR. Within a specific range, urinary Ba (0.56-3.56 µg/L) and urinary Mo (1.06-20.25 µg/L) levels exhibited the most pronounced upwards trend with the risk of IR, while blood Pb (0.05-2.81 µg/dL) and blood Cd (0.24-0.65 µg/L) levels showed a declining trend with IR. The findings on the synergistic effects demonstrated that controlling urinary Ba levels might be more crucial for the management of IR. The SHAP decision plot offered personalized care for IR based on heavy metal control. In conclusion, by utilizing interpretable ML approaches, we emphasize the predictive value of heavy metals for IR, especially Ba, Mo, Pb, and Cd.

Validation of the Framingham Diabetes Risk Model Using Community-Based KoGES Data

BACKGROUND

An 8-year prediction of the Framingham Diabetes Risk Model (FDRM) was proposed, but the predictor has a gap with current clinical standards. Therefore, we evaluated the validity of the original FDRM in Korean population data, developed a modified FDRM by redefining the predictors based on current knowledge, and evaluated the internal and external validity.

METHODS

Using data from a community-based cohort in Korea (n = 5,409), we calculated the probability of diabetes through FDRM, and developed a modified FDRM based on modified definitions of hypertension (HTN) and diabetes. We also added clinical features related to diabetes to the predictive model. Model performance was evaluated and compared by area under the curve (AUC).

RESULTS

During the 8-year follow-up, the cumulative incidence of diabetes was 8.5%. The modified FDRM consisted of age, obesity, HTN, hypo-high-density lipoprotein cholesterol, elevated triglyceride, fasting glucose, and hemoglobin A1c. The expanded clinical model added γ-glutamyl transpeptidase to the modified FDRM. The FDRM showed an estimated AUC of 0.71, and the model's performance improved to an AUC of 0.82 after applying the redefined predictor. Adding clinical features (AUC = 0.83) to the modified FDRM further improved in discrimination, but this was not maintained in the validation data set. External validation was evaluated on population-based cohort data and both modified models performed well, with AUC above 0.82.

CONCLUSION

The performance of FDRM in the Korean population was found to be acceptable for predicting diabetes, but it was improved when corrected with redefined predictors. The validity of the modified model needs to be further evaluated.

Development and validation of machine learning-augmented algorithm for insulin sensitivity assessment in the community and primary care settings: a population-based study in China

Objective

Insulin plays a central role in the regulation of energy and glucose homeostasis, and insulin resistance (IR) is widely considered as the "common soil" of a cluster of cardiometabolic disorders. Assessment of insulin sensitivity is very important in preventing and treating IR-related disease. This study aims to develop and validate machine learning (ML)-augmented algorithms for insulin sensitivity assessment in the community and primary care settings.

Methods

We analyzed the data of 9358 participants over 40 years old who participated in the population-based cohort of the Hubei center of the REACTION study (Risk Evaluation of Cancers in Chinese Diabetic Individuals). Three non-ensemble algorithms and four ensemble algorithms were used to develop the models with 70 non-laboratory variables for the community and 87 (70 non-laboratory and 17 laboratory) variables for the primary care settings to screen the classifier of the state-of-the-art. The models with the best performance were further streamlined using top-ranked 5, 8, 10, 13, 15, and 20 features. Performances of these ML models were evaluated using the area under the receiver operating characteristic curve (AUROC), the area under the precision-recall curve (AUPR), and the Brier score. The Shapley additive explanation (SHAP) analysis was employed to evaluate the importance of features and interpret the models.

Results

The LightGBM models developed for the community (AUROC 0.794, AUPR 0.575, Brier score 0.145) and primary care settings (AUROC 0.867, AUPR 0.705, Brier score 0.119) achieved higher performance than the models constructed by the other six algorithms. The streamlined LightGBM models for the community (AUROC 0.791, AUPR 0.563, Brier score 0.146) and primary care settings (AUROC 0.863, AUPR 0.692, Brier score 0.124) using the 20 top-ranked variables also showed excellent performance. SHAP analysis indicated that the top-ranked features included fasting plasma glucose (FPG), waist circumference (WC), body mass index (BMI), triglycerides (TG), gender, waist-to-height ratio (WHtR), the number of daughters born, resting pulse rate (RPR), etc.

Conclusion

The ML models using the LightGBM algorithm are efficient to predict insulin sensitivity in the community and primary care settings accurately and might potentially become an efficient and practical tool for insulin sensitivity assessment in these settings.

Association of Pooled Fecal Microbiota on Height Growth in Children According to Enterotypes

OBJECTIVES

The association between fecal microbiota and height in children has yielded conflicting findings, warranting further investigation into potential differences in fecal bacterial composition between children with short stature and those of standard height based on enterotypes (ETs).

METHODS

According to the height z score for age and gender, the children were categorized into normal-stature (NS; n = 335) and short-stature (SS; n = 152) groups using a z score of -1.15 as a separator value. The human fecal bacterial FASTA/Q files (n = 487) were pooled and analyzed with the QIIME 2 platform with the National Center for Biotechnology Information alignment search tool. According to ETs, the prediction models by the machine learning algorithms were used for explaining SS, and their quality was validated.

RESULTS

The proportion of SS was 16.4% in ET Enterobacteriaceae (ET-E) and 68.1% in Prevotellaceae (ET-P). The Chao1 and Shannon indexes were significantly lower in the SS than in the NS groups only in ET-P. The fecal bacteria related to SS from the prediction models were similar regardless of ETs. However, in network analysis, the negative correlations between fecal bacteria in the NS and SS groups were much higher in the ET-P than in the ET-E. In the metagenome function, fecal bacteria showed an inverse association of biotin and secondary bile acid synthesis and downregulation of insulin/insulin-like growth factor-1-driven phosphoinositide 3-kinase Akt signaling and AMP-kinase signaling in the SS group compared with the NS group in both ETs.

CONCLUSION

The gut microbial compositions in children were associated with height. Strategies to modify and optimize the gut microbiota composition should be investigated for any potential in promoting height in children.

Interaction of energy and sulfur microbial diet and smoking status with polygenic variants associated with lipoprotein metabolism

Introduction

Hypo-high-density lipoprotein cholesterolemia (hypo-HDL-C) contributes to the development of cardiovascular diseases. The hypothesis that the polygenic variants associated with hypo-HDL-C interact with lifestyle factors was examined in 58,701 middle-aged Korean adults who participated in the Korean Genome and Epidemiology Study (KoGES).

Methods

Participants were categorized into the Low-HDL (case; n = 16,980) and Normal-HDL (n = 41,721) groups. The participants in the Low-HDL group were selected using the guideline-based cutoffs for hypo-HDL-C (<40 mg/dL for men and < 50 mg/dL for women) and included those taking medication for dyslipidemia. The genes associated with hypo-HDL-C were determined through a genome-wide association study (GWAS) in a city hospital-based cohort, and the results were validated in the Ansan/Anung study. The genetic variants for the single nucleotide polymorphism (SNP)-SNP interaction were selected using a generalized multifactor dimensionality reduction analysis, and the polygenic risk score (PRS) generated was evaluated for interaction with lifestyle parameters.

Results

The participants with hypo-HDL-C showed a 1.45 and 1.36-fold higher association with myocardial infarction and stroke, respectively. The High-PRS with four SNPs, namely ZPR1_rs3741297, CETP_rs708272, BUD13_rs180327, and ALDH1A2_rs588136, and that with the 11q23.3 haplotype were positively associated with hypo-HDL-C by about 3 times, which was a 2.4-fold higher association than the PRS of 24 SNP with p < 5×10-8. The risk alleles of CETP_rs708272 and ALDH1A2_rs588136 were linked to increased expression in the heart and decreased in the brain, respectively. The selected SNPs were linked to the reverse cholesterol transport pathway, triglyceride-rich lipoprotein particle remodeling pathway, cholesterol storage, and macrophage-derived foam cell differentiation regulation. The PRS of the 4-SNP model interacted with energy intake and smoking status, while that of the haplotype interacted with a glycemic index of the diet, sulfur microbial diet, and smoking status.

Discussion

Adults with a genetic risk for hypo-HDL-C need to modulate their diet and smoking status to reduce their risk.

Diabetes Mellitus in Pancreatic Cancer: A Distinct Approach to Older Subjects with New-Onset Diabetes Mellitus

BACKGROUND

Pancreatic ductal adenocarcinoma (PDAC) is associated with a very poor prognosis, with near-identical incidence and mortality. According to the World Health Organization Globocan Database, the estimated number of new cases worldwide will rise by 70% between 2020 and 2040. There are no effective screening methods available so far, even for high-risk individuals. The prognosis of PDAC, even at its early stages, is still mostly unsatisfactory. Impaired glucose metabolism is present in about 3/4 of PDAC cases.

METHODS

Available literature on pancreatic cancer and diabetes mellitus was reviewed using a PubMed database. Data from a national oncology registry (on PDAC) and information from a registry of healthcare providers (on diabetes mellitus and a number of abdominal ultrasound investigations) were obtained.

RESULTS

New-onset diabetes mellitus in subjects older than 60 years should be an incentive for a prompt and detailed investigation to exclude PDAC. Type 2 diabetes mellitus, diabetes mellitus associated with chronic non-malignant diseases of the exocrine pancreas, and PDAC-associated type 3c diabetes mellitus are the most frequent types. Proper differentiation of particular types of new-onset diabetes mellitus is a starting point for a population-based program. An algorithm for subsequent steps of the workup was proposed.

CONCLUSIONS

The structured, well-differentiated, and elaborately designed approach to the elderly with a new onset of diabetes mellitus could improve the current situation in diagnostics and subsequent poor outcomes of therapy of PDAC.

Polygenic Variants Linked to Oxidative Stress and the Antioxidant System Are Associated with Type 2 Diabetes Risk and Interact with Lifestyle Factors

Oxidative stress is associated with insulin resistance and secretion, and antioxidant systems are essential for preventing and managing type 2 diabetes (T2DM). This study aimed to explore the polygenic variants linked to oxidative stress and the antioxidant system among those associated with T2DM and the interaction of their polygenic risk scores (PRSs) with lifestyle factors in a large hospital-based cohort (n = 58,701). Genotyping, anthropometric, biochemical, and dietary assessments were conducted for all participants with an average body mass index of 23.9 kg/m². Genetic variants associated with T2DM were searched through genome-wide association studies in participants with T2DM (n = 5383) and without T2DM (n = 53,318). The Gene Ontology database was searched for the antioxidant systems and oxidative stress-related genes among the genetic variants associated with T2DM risk, and the PRS was generated by summing the risk alleles of selected ones. Gene expression according to the genetic variant alleles was determined on the FUMA website. Food components with low binding energy to the GSTA5 protein generated from the wildtype and mutated GSTA5_rs7739421 (missense mutation) genes were selected using in silico analysis. Glutathione metabolism-related genes, including glutathione peroxidase (GPX)1 and GPX3, glutathione disulfide reductase (GSR), peroxiredoxin-6 (PRDX6), glutamate-cysteine ligase catalytic subunit (GCLC), glutathione S-transferase alpha-5 (GSTA5), and gamma-glutamyltransferase-1 (GGT1), were predominantly selected with a relevance score of >7. The PRS related to the antioxidant system was positively associated with T2DM (ORs = 1.423, 95% CI = 1.22-1.66). The active site of the GASTA proteins having valine or leucine at 55 due to the missense mutation (rs7739421) had a low binding energy (<-10 kcal/mol) similarly or differently to some flavonoids and anthocyanins. The PRS interacted with the intake of bioactive components (specifically dietary antioxidants, vitamin C, vitamin D, and coffee) and smoking status (p < 0.05). In conclusion, individuals with a higher PRS related to the antioxidant system may have an increased risk of T2DM, and there is a potential indication that exogenous antioxidant intake may alleviate this risk, providing insights for personalized strategies in T2DM prevention.

Fecal Microbiota Composition, Their Interactions, and Metagenome Function in US Adults with Type 2 Diabetes According to Enterotypes

T2DM etiology differs among Asians and Caucasians and may be associated with gut microbiota influenced by different diet patterns. However, the association between fecal bacterial composition, enterotypes, and T2DM susceptibility remained controversial. We investigated the fecal bacterial composition, co-abundance network, and metagenome function in US adults with T2DM compared to healthy adults based on enterotypes. We analyzed 1911 fecal bacterial files of 1039 T2DM and 872 healthy US adults from the Human Microbiome Projects. Operational taxonomic units were obtained after filtering and cleaning the files using Qiime2 tools. Machine learning and network analysis identified primary bacteria and their interactions influencing T2DM incidence, clustered into enterotypes, Bacteroidaceae (ET-B), Lachnospiraceae (ET-L), and Prevotellaceae (ET-P). ET-B showed higher T2DM incidence. Alpha-diversity was significantly lower in T2DM in ET-L and ET-P (p < 0.0001), but not in ET-B. Beta-diversity revealed a distinct separation between T2DM and healthy groups across all enterotypes (p < 0.0001). The XGBoost model exhibited high accuracy and sensitivity. Enterocloster bolteae, Facalicatena fissicatena, Clostridium symbiosum, and Facalibacterium prausnitizii were more abundant in the T2DM group than in the healthy group. Bacteroides koreensis, Oscillibacter ruminantium, Bacteroides uniformis, and Blautia wexlerae were lower in the T2DM than in the healthy group regardless of the enterotypes in the XGBoost model (p < 0.0001). However, the patterns of microbial interactions varied among different enterotypes affecting T2DM risk. The interaction between fecal bacteria was more tightly regulated in the ET-L than in the ET-B and ET-P groups (p < 0.001). Metagenomic analysis revealed an inverse association between bacteria abundance in T2DM, energy utility, butanoate and propanoate metabolism, and the insulin signaling pathway (p < 0.0001). In conclusion, fecal bacteria play a role in T2DM pathogenesis, particularly within different enterotypes, providing valuable insights into the link between gut microbiota and T2DM in the US population.

A Prediction Model for Osteoporosis Risk Using a Machine-Learning Approach and Its Validation in a Large Cohort

BACKGROUND

Osteoporosis develops in the elderly due to decreased bone mineral density (BMD), potentially increasing bone fracture risk. However, the BMD is not regularly measured in a clinical setting. This study aimed to develop a good prediction model for the osteoporosis risk using a machine learning (ML) approach in adults over 40 years in the Ansan/Anseong cohort and the association of predicted osteoporosis risk with a fracture in the Health Examinees (HEXA) cohort.

METHODS

The 109 demographic, anthropometric, biochemical, genetic, nutrient, and lifestyle variables of 8,842 participants were manually selected in an Ansan/Anseong cohort and included in the ML algorithm. The polygenic risk score (PRS) of osteoporosis was generated with a genome-wide association study and added for the genetic impact of osteoporosis. Osteoporosis was defined with < -2.5 T scores of the tibia or radius compared to people in their 20s-30s. They were divided randomly into the training (n = 7,074) and test (n = 1,768) sets-Pearson's correlation between the predicted osteoporosis risk and fracture in the HEXA cohort.

RESULTS

XGBoost, deep neural network, and random forest generated the prediction model with a high area under the curve (AUC, 0.86) of the receiver operating characteristic (ROC) with 10, 15, and 20 features; the prediction model by XGBoost had the highest AUC of ROC, high accuracy and k-fold values (> 0.85) in 15 features among seven ML approaches. The model included the genetic factor, genders, number of children and breastfed children, age, residence area, education, seasons to measure, height, smoking status, hormone replacement therapy, serum albumin, hip circumferences, vitamin B6 intake, and body weight. The prediction models for women alone were similar to those for both genders, with lower accuracy. When the prediction model was applied to the HEXA study, the correlation between the fracture incidence and predicted osteoporosis risk was significant but weak (r = 0.173, P < 0.001).

CONCLUSION

The prediction model for osteoporosis risk generated by XGBoost can be applied to estimate osteoporosis risk. The biomarkers can be considered for enhancing the prevention, detection, and early therapy of osteoporosis risk in Asians.

Height-Related Polygenic Variants Are Associated with Metabolic Syndrome Risk and Interact with Energy Intake and a Rice-Main Diet to Influence Height in KoGES

Adult height is inversely related to metabolic syndrome (MetS) risk, but its genetic impacts have not been revealed. The present study aimed to examine the hypothesis that adult height-related genetic variants interact with lifestyle to influence adult height and are associated with MetS risk in adults aged >40 in Korea during 2010–2014. Participants were divided into short stature (SS; control) and tall stature (TS; case) by the 85th percentile of adult height. The genetic variants linked to adult height were screened from a genome-wide association study in a city hospital-based cohort (n = 58,701) and confirmed in Ansan/Ansung plus rural cohorts (n = 13,783) among the Korean Genome and Epidemiology Study. Genetic variants that interacted with each other were identified using the generalized multifactor dimensionality reduction (GMDR) analysis. The interaction between the polygenic risk score (PRS) of the selected genetic variants and lifestyles was examined. Adult height was inversely associated with MetS, cardiovascular diseases, and liver function. The PRS, including zinc finger and BTB domain containing 38 (ZBTB38)_rs6762722, polyadenylate-binding protein-interacting protein-2B (PAIP2B)_rs13034890, carboxypeptidase Z (CPZ)_rs3756173, and latent-transforming growth factor beta-binding protein-1 (LTBP1)_rs4630744, was positively associated with height by 1.29 times and inversely with MetS by 0.894 times after adjusting for covariates. In expression quantitative trait loci, the gene expression of growth/differentiation factor-5 (GDF5)_rs224331, non-SMC condensin I complex subunit G (NCAPG)_rs2074974, ligand-dependent nuclear receptor corepressor like (LCORL)_rs7700107, and insulin-like growth factor-1 receptor (IGF1R)_rs2871865 was inversely linked to their risk allele in the tibial nerve and brain. The gene expression of PAIP2B_rs13034890 and a disintegrin and metalloproteinase with thrombospondin motifs-like-3 (ADAMTSL3)_rs13034890 was positively related to it. The PRS was inversely associated with MetS, hyperglycemia, HbA1c, and white blood cell counts. The wild type of GDF5_rs224331 (Ala276) lowered binding energy with rugosin A, D, and E (one of the hydrolyzable tannins) but not the mutated one (276Ser) in the in-silico analysis. The PRS interacted with energy intake and rice-main diet; PRS impact was higher in the high energy intake and the low rice-main diet. In conclusion, the PRS for adult height interacted with energy intake and diet patterns to modulate height and was linked to height and MetS by modulating their expression in the tibial nerve and brain.

Development and validation of an insulin resistance model for a population without diabetes mellitus and its clinical implication: a prospective cohort study

Background

Insulin resistance (IR) is associated with diabetes mellitus, cardiovascular disease (CV), and mortality. Few studies have used machine learning to predict IR in the non-diabetic population.

Methods

In this prospective cohort study, we trained a predictive model for IR in the non-diabetic populations using the US National Health and Nutrition Examination Survey (NHANES, from JAN 01, 1999 to DEC 31, 2012) database and the Taiwan MAJOR (from JAN 01, 2008 to DEC 31, 2017) database. We analysed participants in the NHANES and MAJOR and participants were excluded if they were aged <18 years old, had incomplete laboratory data, or had DM. To investigate the clinical implications (CV and all-cause mortality) of this trained model, we tested it with the Taiwan biobank (TWB) database from DEC 10, 2008 to NOV 30, 2018. We then used SHapley Additive exPlanation (SHAP) values to explain differences across the machine learning models.

Findings

Of all participants (combined NHANES and MJ databases), we randomly selected 14,705 participants for the training group, and 4018 participants for the validation group. In the validation group, their areas under the curve (AUC) were all >0.8 (highest being XGboost, 0.87). In the test group, all AUC were also >0.80 (highest being XGboost, 0.88). Among all 9 features (age, gender, race, body mass index, fasting plasma glucose (FPG), glycohemoglobin, triglyceride, total cholesterol and high-density cholesterol), BMI had the highest value of feature importance on IR (0.43 for XGboost and 0.47 for RF algorithms). All participants from the TWB database were separated into the IR group and the non-IR group according to the XGboost algorithm. The Kaplan-Meier survival curve showed a significant difference between the IR and non-IR groups (p < 0.0001 for CV mortality, and p = 0.0006 for all-cause mortality). Therefore, the XGboost model has clear clinical implications for predicting IR, aside from CV and all-cause mortality.

Interpretation

To predict IR in non-diabetic patients with high accuracy, only 9 easily obtained features are needed for prediction accuracy using our machine learning model. Similarly, the model predicts IR patients with significantly higher CV and all-cause mortality. The model can be applied to both Asian and Caucasian populations in clinical practice.

Funding

Taichung Veterans General Hospital, Taiwan and Japan Society for the Promotion of Science KAKENHI Grant Number JP21KK0293.

Association of Polygenic Variants Involved in Immunity and Inflammation with Duodenal Ulcer Risk and Their Interaction with Irregular Eating Habits

Genetic and environmental factors are associated with developing and progressing duodenal ulcer (DU) risk. However, the exact nature of the disease pathophysiology and the single nucleotide polymorphism (SNP)-lifestyle interaction has yet to be determined. The purpose of the present study was to examine the SNPs linked to DU risk and their interaction with lifestyles and diets in a large hospital-based cohort of Asians. Based on an earlier diagnosis, the participants were divided into the DU (case; n = 1088) and non-DU (control, n = 56,713) groups. The SNP associated with DU risk were obtained from a genome-wide association study (GWAS), and those promoted genetic impact with SNP-SNP interactions were identified with generalized multifactor dimensionality reduction analysis. The interaction between polygenic risk score (PRS) calculated from the selected genetic variants and nutrient were examined. They were related to actin modification, immune response, and cell migration by modulating leucine-rich repeats (LRR) domain binding, Shaffer interferon regulatory factor 4 (IRF4) targets in myeloma vs. mature B lymphocyte, and Reactome runt-related transcription factor 3 (RUNX3). Among the selected SNPs, rs11230563 (R225W) showed missense mutation and low binding affinity with different food components in the autodock analysis. Glycyrrhizin, physalin B, janthitrem F, and casuarinin lowered it in only wild CD6 protein but not in mutated CD6. Plastoquinone 8, solamargine, saponin D, and matesaponin 2 decreased energy binding affinity in mutated CD6 proteins. The PRS of the 5-SNP and 6-SNP models exhibited a positive association with DU risk (OR = 3.14). The PRS of the 5-SNP PRS model interacted with irregular eating habits and smoking status. In participants with irregular eating habits or smokers, DU incidence was much higher in the participants with high PRS than in those with low PRS. In conclusion, the genetic impact of DU risk was mainly in regulating immunity, inflammation, and actin modification. Adults who are genetically susceptible to DU need to eat regularly and to be non-smokers. The results could be applied to personalize nutrition.

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.

Fecal Bacterial Community and Metagenome Function in Asians with Type 2 Diabetes, According to Enterotypes

The role of gut microbes has been suggested in type 2 diabetes (T2DM) risk. However, their results remain controversial. We hypothesized that Asians with T2DM had different fecal bacterial compositions, co-abundance networks, and metagenome functions compared to healthy individuals, according to enterotypes. This hypothesis was examined using the combined gut microbiota data from human fecal samples from previous studies. The human fecal bacterial FASTA/Q files from 36 different T2DM studies in Asians were combined (healthy, n = 3378; T2DM, n = 551), and operational taxonomic units (OTUs) and their counts were obtained using qiime2 tools. In the machine learning approaches, fecal bacteria rich in T2DM were found. They were separated into two enterotypes, Lachnospiraceae (ET-L) and Prevotellaceae (ET-P). The Shannon and Chao1 indices, representing α-diversity, were significantly lower in the T2DM group compared to the healthy group in ET-L (p < 0.05) but not in ET-P. In the Shapley additive explanations analysis of ET-L, Escherichia fergusonii, Collinsella aerofaciens, Streptococcus vestibularis, and Bifidobacterium longum were higher (p < 0.001), while Phocaeicola vulgatus, Bacteroides uniformis, and Faecalibacterium prausnitzii were lower in the T2DM group than in the healthy group (p < 0.00005). In ET-P, Escherichia fergusonii, Megasphaera elsdenii, and Oscillibacter valericigenes were higher, and Bacteroides koreensis and Faecalibacterium prausnitzii were lower in the T2DM group than in the healthy group. In ET-L and ET-P, bacteria in the healthy and T2DM groups positively interacted with each other within each group (p < 0.0001) but negatively interacted between the T2DM and healthy groups in the network analysis (p < 0.0001). In the metagenome functions of the fecal bacteria, the gluconeogenesis, glycolysis, and amino acid metabolism pathways were higher, whereas insulin signaling and adenosine 5′ monophosphate-activated protein kinase (AMPK) signaling pathways were lower in the T2DM group than in the healthy group for both enterotypes (p < 0.00005). In conclusion, Asians with T2DM exhibited gut dysbiosis, potentially linked to intestinal permeability and the enteric vagus nervous system.

Association of Polygenic Variants with Type 2 Diabetes Risk and Their Interaction with Lifestyles in Asians

Over the last several decades, there has been a considerable growth in type 2 diabetes (T2DM) in Asians. A pathophysiological mechanism in Asian T2DM is closely linked to low insulin secretion, β-cell mass, and inability to compensate for insulin resistance. We hypothesized that genetic variants associated with lower β-cell mass and function and their combination with unhealthy lifestyle factors significantly raise T2DM risk among Asians. This hypothesis was explored with participants aged over 40. Participants were categorized into T2DM (case; n = 5383) and control (n = 53,318) groups. The genetic variants associated with a higher risk of T2DM were selected from a genome-wide association study in a city hospital-based cohort, and they were confirmed with a replicate study in Ansan/Ansung plus rural cohorts. The interacted genetic variants were identified with generalized multifactor dimensionality reduction analysis, and the polygenic risk score (PRS)-nutrient interactions were examined. The 8-SNP model was positively associated with T2DM risk by about 10 times, exhibiting a higher association than the 20-SNP model, including all T2DM-linked SNPs with p < 5 × 10−6. The SNPs in the models were primarily involved in pancreatic β-cell growth and survival. The PRS of the 8-SNP model interacted with three lifestyle factors: energy intake based on the estimated energy requirement (EER), Western-style diet (WSD), and smoking status. Fasting serum glucose concentrations were much higher in the participants with High-PRS in rather low EER intake and high-WSD compared to the High-EER and Low-WSD, respectively. They were shown to be higher in the participants with High-PRS in smokers than in non-smokers. In conclusion, the genetic impact of T2DM risk was mainly involved with regulating pancreatic β-cell mass and function, and the PRS interacted with lifestyles. These results highlight the interaction between genetic impacts and lifestyles in precision nutrition.

Simple Method to Predict Insulin Resistance in Children Aged 6-12 Years by Using Machine Learning

BACKGROUND

Due to the increasing insulin resistance (IR) in childhood, rates of diabetes and cardiovascular disease may rise in the future and seriously threaten the healthy development of children. Finding an easy way to predict IR in children can help pediatricians to identify these children in time and intervene appropriately, which is particularly important for practitioners in primary health care.

PATIENTS AND METHODS

Seventeen features from 503 children 6-12 years old were collected. We defined IR by HOMA-IR greater than 3.0, thus classifying children with IR and those without IR. Data were preprocessed by multivariate imputation and oversampling to resolve missing values and data imbalances; then, recursive feature elimination was applied to further select features of interest, and 5 machine learning methods-namely, logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), and gradient boosting with categorical features support (CatBoost)-were used for model training. We tested the trained models on an external test set containing information from 133 children, from which performance metrics were extracted and the optimal model was selected.

RESULTS

After feature selection, the numbers of chosen features for the LR, SVM, RF, XGBoost, and CatBoost models were 6, 9, 10, 14, and 6, respectively. Among them, glucose, waist circumference, and age were chosen as predictors by most of the models. Finally, all 5 models achieved good performance on the external test set. Both XGBoost and CatBoost had the same AUC (0.85), which was highest among those of all models. Their accuracy, sensitivity, precision, and F1 scores were also close, but the specificity of XGBoost reached 0.79, which was significantly higher than that of CatBoost, so XGBoost was chosen as the optimal model.

CONCLUSION

The model developed herein has a good predictive ability for IR in children 6-12 years old and can be clinically applied to help pediatricians identify children with IR in a simple and inexpensive way.