Metabolic Fingerprints of Diabetes: Machine Learning Reveals Distinct Biomarkers in Type 2 Diabetes Mellitus

Metabolic Fingerprints of Diabetes: Machine Learning Reveals Distinct Biomarkers in Type 2 Diabetes Mellitus | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Metabolic Fingerprints of Diabetes: Machine Learning Reveals Distinct Biomarkers in Type 2 Diabetes Mellitus Mohammad T. Al Bataineh, Andre Sanches Barreiros, Joviana Farhat, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6606669/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Type 2 Diabetes Mellitus (T2DM) and obesity have reached epidemic levels globally. Although T2DM strain a heavy burden on population health, limited information is known about its metabolic signature, particularly in Emirati population. We explored Emiratis with controlled and uncontrolled T2DM metabolic profiles employing advanced machine-learning methods e.g. predictor randomization and bootstrap aggregation. Alpha-ketoisovaleric acid, 2-pyrrolidinone, and uridine were identified as significant metabolic markers related to inflammation and metabolic stress between the two groups. Notable interactions between metabolites in the tryptophan metabolic pathway, including L-tryptophan and indole compounds, were observed, highlighting their potential role in T2DM progression. Alterations in Tricarboxylic Acid cycle intermediates and increased activation of Pentose Phosphate Pathway suggested adaptive responses to oxidative stress. Furthermore, metabolic changes were observed across the prediabetic, controlled, and uncontrolled diabetes stages, with metabolites such as 5-Dodecenoic acid and phosphatidylcholine identified as potential markers for distinguishing between these stages. Our novel findings reveal the complex metabolic alterations associated with T2DM, providing an in-depth insight into the molecular level. These insights suggest that specific metabolites could serve as reliable biomarkers for early diagnosis and personalized treatment strategies. This approach could revolutionize the chronic conditions management, with effective and tailored interventions. Health sciences/Endocrinology Health sciences/Medical research Health sciences/Molecular medicine Diabetes Mellitus Inflammation Metabolite Machine Learning Figures Figure 1 Figure 2 Figure 3 Introduction Metabolic disorders have been a major public health issue and a clinical challenge worldwide 1 . Type 2 Diabetes Mellitus (T2DM) is one of the most common metabolic diseases that affects approximately 422–537 million people worldwide 2 . In 2021, it was estimated that approximately 6.7 million deaths were caused by T2DM and its complications 2 , making it one of the top ten causes of mortality globally 3 . At a regional level, the Middle East and North Africa region (MENA) had a high prevalence of DM in 2019 (12.2%), which is expected to increase by 96% between 2019 and 2045 4 . In clinical practice, T2DM is the most diagnosed type of diabetes, accounting for approximately 90% of all cases 5 . T2DM is characterized by a gained physiological insulin resistance due to islet cell dysfunction 6 . The persistent increase in the burden of T2DM and its complications highlights the need for a deeper understanding of the underlying metabolic disturbances. Therefore, identifying biomarkers during the screening and prediction stages can allow for more personalized healthcare management 7 . In practice, multiple biomarkers, including fasting plasma glucose (FPG), glycated hemoglobin A1c (HbA1c), triglycerides, HDL cholesterol, inflammatory mediators, adiponectin, liver enzymes, and fetuin-A have been used for estimating T2DM risk 8 . However, many of these biomarkers do not fully capture the complexity of T2DM etiology 9 . Therefore, metabolomics has been applied to identify and quantify metabolites as they can better reflect physiological dysfunctions and allow the earlier detection of T2DM 10 . Metabolomics studies have revealed several blood sugars, sugar-related metabolites, glycolysis/gluconeogenesis pathway components, and tricarboxylic acid (TCA) cycle intermediates associated with diabetes 11 . Other studies revealed novel biomarkers associated with T2DM, such as blood acylcarnitines, phenylalanine metabolites, energy metabolism, and lipid metabolism 12 . Branched-chain amino acids (BCAA) such as valine, leucine, isoleucine, and phospholipids were among the most popular biomarkers associated with T2DM progression 13 . Several molecular pathways are also involved in the pathophysiology of T2DM through the induction of oxidative stress 14 . Accordingly, alterations in these signaling pathways have been linked to disease progression 15 . The limited understanding of the specific molecular mechanisms and epidemiological status underlying T2DM in the MENA region may hinder effective treatment strategies 16 . Both genetic and environmental factors play an important role in determining an individual's risk of developing T2DM 17 . Identifying genetic variations associated with increased disease risk could be essential for improving disease management, enhancing clinical outcomes, and preventing complications 17 . In practice, evaluating the intrinsic pathways during the course of T2DM is needed for non-invasive diagnosis and long-term targeted therapy 15 . This is particularly relevant to patients in the United Arab Emirates (UAE) who are at high risk of T2DM and its complications 18 . Thus, our study aims to analyze preliminary data from Emirati diabetic patients with uncontrolled and controlled disease status using more advanced methodologies. Specifically, it employs univariate analysis of covariant alongside the powerful machine-learning AI tool, Random Forest (RF). As a regression tree technique, RF uses bootstrap aggregation and predictor randomization to achieve high predictive accuracy. This approach is particularly effective in handling complex data-generating processes to ensure optimal model performance. Materials and Methods Study Participants A case-control study was conducted on adult patients recruited from the Diabetes and Endocrine Center at the University Hospital Sharjah in the UAE. The cohort was defined as patients clinically diagnosed with T2DM and with a Hemoglobin Subunit Alpha 1 (HbA1) ≥ 5.7%. Participants with an HbA1c ≥ 5.7% but were not clinically confirmed to be diabetic were chosen as controls. The approval for the study conduction was obtained from the hospital’s Research and Ethics Committee. All experiments were performed in accordance with relevant guidelines and regulations. Sample collection, preparation, and analysis Blood specimens were carefully procured from the patients, each of whom had given informed consent and underwent the sampling procedure in a controlled clinical setting under the supervision of trained medical personnel. In the current study, the protocols for sample preparation and analysis adhered to the methodologies previously employed in a similar study 19 . Statistical analysis The patient's data was reanalyzed using statistical tools on R version 4.2.2. The metabolic concentration values were normalized as shown in MetaboAnalyst® 5.0, where data was first cleaned, then standardized and normalized through the logarithmic transformation. To address the age bias between the cohorts, a conditional linear regression was employed based on the provided data 20 , 21 . The regression was performed under the condition that the following conditions were met: Primarily, the gradient of the best-fit line or the control patients (m c ) had to be within 2 standard deviations from the gradient of the best-fit line for all patients (m a ). Subsequently, the mean metabolite value for control and case patients had to be at least 2 standard deviations apart. Finally, the correlation coefficient for the control patients is greater than |0.3|. These conditions exist to ensure that only the metabolites that show a linear correlation with age would be corrected. Otherwise, some correction artifacts may arise due to the strong correlation between the two variables, age and T2DM. Random Forest (RF) in machine learning (ML) The RF machine learning algorithm from the R library was implemented afterward to identify a ranked list of metabolites contributing to the separation between both groups. RF is a type of general-purpose classification and regression ML method that can be performed in settings where the number of variables exceeds the number of observables 22 . The algorithm combines several randomized decision trees in the learning phase and averages their predictions to obtain a final result 23 . The RF method applied a pseudo-random division of the patients’ data set into training and test sets. A seed ranging from 1 to 100 (all-inclusive) split the data into two sets. The training set comprised 80% of the patients and was used to ‘teach’ the algorithm to distinguish between the different groups based on their 148 metabolites. The RF algorithm is then tested on the test set (20% of the data) by attempting to categorize the patients based on their metabolites correctly. Each run weighs the metabolite effects differently depending on the training data set. Therefore, it is necessary to average the resulting metabolites throughout the runs. This was done by averaging the Mean Decrease Gini for each metabolite across all the runs, from which an ordered list of the most significant metabolites can be generated. Covariance analytical method In addition to the RF method described above, a covariance analysis was performed to further quantify the significant metabolites distinguishing between the two patient cohorts. A one-way ANOVA test compared the normalized metabolite concentration between the controls and the cases. A metabolite was considered significantly correlated if the linear correlation between its concentrations and the patient cohort had a False Discovery Rate (FDR) p-value < 0.05. This method quantifies the linear dependence for each metabolite on the cohorts, and the strength of this correlation is recorded. A secondary aspect investigated with the linear covariance method was the correlation between different metabolites. Strong correlations (|R| >0.8) would indicate metabolites that can be grouped together, suggesting involvement in similar metabolic pathways. A tertiary aspect of this method was identifying metabolites that strongly correlated with patients having 5.7% ≤ HbA1c ≤ 6.4% and HbA1c ≥ 6.5%. These metabolites could help distinguish between prediabetic/controlled and uncontrolled diabetics. Pathway analysis A combined list of significant metabolites from the RF and the covariant methods was created, followed by pathway analysis. Pathway analysis was performed using the online tool MetaboAnalyst® 5.0 24 . The analysis was done with the following settings: the enrichment method was set to a scatter plot visualization method, the Hypergeometric test was applied, the topological analysis used relative betweenness Centrality, and the Homo sapiens Kyoto Encyclopedia of Genes and Genome (KEGG) pathway library was used. The resulting figure displays the pathways in which the metabolites are involved. The final step in this study was to combine the pathways from the pathway analysis into a super pathway. This was done by linking the most significant pathways from the analysis through the key metabolites identified by the RF and covariant methods, in conjunction with the KEGG library. A table listing the metabolites was provided, showing the fold Log2 change and the FDR q-value of the metabolite. A positive fold Log2 indicates an amplified metabolite in the cases cohort or a suppressed metabolite in the control cohort, while a negative fold Log2 indicates the opposite. Results Study participants Seventy-six subjects were voluntarily enrolled in this study; 50 were diagnosed with T2DM and treated at the University Hospital Sharjah. None of the other 26 patients had been diagnosed with T2DM and were considered the control group in this study. The evaluation of study participants resulted in significant differences between case and control groups that are relevant to detecting T2DM biomarkers. Most cases were older than 65 years, while most controls were under 65 years, indicating a strong association between age and T2DM progression (p = 4.10 x 10⁻⁸). Additionally, a higher proportion of cases had a BMI > 25 kg/m² compared to controls, yielding a borderline significant difference in BMI between both groups (p = 0.050). HbA1c levels were also significantly different between the groups (p = 0.024), with more cases having HbA1c levels ≥ 6.5%, while a higher percentage of controls fell within the pre-diabetic range (5.7% ≤ HbA1c ≤ 6.4%) (See Table 1 ). Table 1 The characteristics of study participants Variables Cases (n = 50) Controls (n = 26) p-value n (%) n (%) Gender 0.945 Male 15 (30.0%) 8 (30.8%) Female 35 (70.0%) 18 (69.2%) Age 4.10 x 10 − 8 Age 25 kg/m2 34 (68.0%) 18 (69.2%) N/A (Wheelchair) 8 (16.0%) 0 (0.0%) Blood sugar level (HbA1c) 5.7% ≤ HbA1c ≤ 6.4% 23 (46.0%) 19 (73.1%) 0.024 HbA1c ≥ 6.5% 27 (54.0%) 7 (26.9%) Random Forest (RF) in machine learning (ML) The first method used to identify the metabolites that distinguish the two groups (diagnosed vs undiagnosed T2DM) is the RF algorithm from the R library 25 . This method ranks metabolites based on their importance in categorizing patients, producing a list from the most to the least significant. Figure 1 A & B shows an example of this for a particular run. Figure 2 A & B shows that for seed 100, the RF method identifies Alpha-ketoisovaleric acid, 2-pyrrolidinone, and Uridine as the most significant metabolites, respectively. However, different runs may rank the metabolites differently, as the order highly depends on how the seed splits the training and testing set. The RF method can correctly classify patients as not diagnosed vs undiagnosed with T2DM, with an average accuracy of 88.0% across 1000 runs. Covariance analytical method Alpha-ketoisovaleric acid is the most significant metabolite for the covariance method. In addition to the RF method, the covariance (using one-way ANOVA with FDR q-value < 0.05) was applied to assess the relationship between metabolites and diagnosis. The covariance method complements the RF method by determining correlations and providing p-values for each metabolite. The results are shown in Supplementary Table 1. Supplementary Table 1 lists the metabolites Alpha-ketoisovaleric acid, N-Methylhydantoin, 3-Methylindole, Indole-3-carbinol and L-Tryptophan as the most significant for the one-way ANOVA test. This result aligns with Fig. 1 , where these same metabolites were also identified as the most significant by the RF method. Alpha-ketoisovaleric acid, N-Methylhydantoin, Indole-3-carbinol, L-Tryptophan, and Indole are all strongly (positively) correlated with each other, with R > 0.9 (Supplementary Fig. 1). Together, these five metabolites are the most significant in identifying patient groupings. These five metabolites (except for Alpha-ketoisovaleric acid) and Indoleacetic acid, 5-Methoxytryptophol, and Indolelactic acid are all involved in the Tryptophan metabolism 27 . However, unlike metabolites like N-Methylhydantoin, these metabolites are overexpressed in the cases of patients. The metabolites 5-Dodecenoic acid, 4-Ethylbenzoic acid, Succinylacetone, Hexadecanedioic acid, 2-Phenylbutyric acid, Homovanillic acid, and PC (18:1(9Z)/18:1(9Z)) distinguish prediabetic/controlled diabetic from uncontrolled diabetic patients, as shown in Supplementary Fig. 2. These metabolites are also strongly (positively) correlated (> 0.8) with each other and may be influenced by the blood sugar level bias observed in the data in Table 1 . Pathway analysis The RF and covariance methods have identified significant metabolites distinguishing between diagnosed and undiagnosed diabetic patients. The next step is to identify the most significant metabolic pathways involving these metabolites through the pathway analysis method in MetaboAnalyst 5.0. All metabolites listed in Fig. 1 and Supplementary Table 1 were processed using the KEGG Homo sapiens pathway library 28 to find the significant metabolic pathways. The resulting pathways are shown in Fig. 2 . As shown in Fig. 2 A, the most impactful pathways with a fold change > 1 include Caffeine metabolism, Glyoxylate and Dicarboxylate metabolism, Aminoacyl-tRNA biosynthesis, TCA cycle, Linoleic acid metabolism, Tryptophan metabolism, Valine, leucine and isoleucine biosynthesis, and Vitamin B6 metabolism. As shown in Fig. 2 B, the pathways, Caffeine metabolism, and valine, leucine, and isoleucine, biosynthesis have a p-value > 0.05 and thus are not considered significant. The table also shows that Linoleic acid metabolism and Aminoacyl-tRNA biosynthesis have zero impact and are insignificant. The remaining pathways were considered when constructing the combined superpathway. Superpathway construction Using the most significant pathways from Fig. 2 and metabolites from Fig. 1 and Supplementary Table 1, a link between these can be established into one superpathway to provide a possible metabolic picture of the biological processes at work. This is represented in Fig. 3 A & B as a network, with all connections made using the KEGG library 28 . The superpathway created shows that the significant pathways from Fig. 2 are connected through the Glycolysis pathway. It begins with Glucose (which is not a metabolite in the human body) and converts it into Alpha-ketoisovaleric acid after several intermediate metabolites. Some of these intermediate metabolites branch off into Pyridoxal 5’-phosphate via the Vitamin B6 metabolism pathway. Later intermediate metabolites can be converted into Glycine or L-Tryptophan instead, continuing down the Tryptophan metabolism pathway. As shown on the Table in Fig. 3 B, the end products of this pathway, including Indole-3-carbinol, Indole, and 3-methylindole, have a negative fold-change, indicating a higher concertation in the control cohort. Conversely, the pathway may lead to metabolites such as 5-methoxytryptophol, N-Acetylserotonin, Indolelactic acid, and Indoleacetic acid, with a positive fold-change, indicating higher concentrations in the case cohort. The final metabolite of glycolysis, Alpha-ketoisovaleric acid, can then enter the TCA cycle via citric acid, which is converted into cis-aconitic acid. In the reverse direction, cis-aconitic acid can convert into citric acid, as part of the Glyoxylate and dicarboxylate metabolism pathway. Discussion We investigated data from diabetic patients with controlled and uncontrolled T2DM utilizing sophisticated approaches, such as ML methods, bootstrap aggregation, and randomizing predictors to maximize predictive accuracy. The evaluation of study participants revealed differences in age, BMI, and HBA1c levels between cases and controls. Prior studies showed a strong association between higher BMI and increased risk of T2DM, in which obesity constitutes a critical factor for disease development 29 . However, a more complex relationship between BMI and diabetes risk was previously reported, suggesting that factors such as muscle mass and distribution of fat may also play significant roles in disease progression 30 . This highlights the importance of BMI and glycemic control as key factors in understanding the metabolic differences between cases and controls, reinforcing their relevance when assessing novel biomarkers in T2DM. First, we explored the significantly expressed metabolic markers in diabetic subgroups. The study shows that alpha-ketoisovaleric acid, 2-Pyrrolidinone, and uridine are the most significantly expressed metabolic markers in patients with a confirmed T2DM diagnosis. Valine, Leucine, and Isoleucine biosynthesis were mainly upregulated in the diabetic group. This aligns with existing evidence reporting the association between elevated levels of alpha-ketoisovaleric acid and the progression of metabolic disorders, including diabetes 31 . Therefore, accumulating branched-chain amino acids (BCAA) intermediates such as alpha-ketoisovaleric acid may contribute to metabolic stress and inflammation that can exacerbate diabetes progression 32 . A recent study highlighted the link between increased circulating BCAA metabolites, including alpha-ketoisovaleric acid, and the risk of developing insulin resistance and T2DM 33 . It has also been suggested that uridine can influence insulin signaling 34 , with potential implications for glucose homeostasis in T2DM. Urasaki et al. reported that uridine was shown to regulate glucose homeostasis by playing a dual role, as both high and low levels can affect insulin sensitivity 35 , presenting uridine as a potential modulator in metabolic diseases like T2DM. Recently, fasting uridine was found to be closely associated with carotid atherosclerosis in patients with T2DM, constituting a promising predictor of diabetes 36 . Next, to explore the relevant correlation of metabolic pathways in diabetic subgroups, we determined that Alpha-ketoisovaleric acid, N-Methylhydantoin, Indole-3-carbinol, L-Tryptophan, and Indole exhibit strong positive correlations. These metabolites were highly interrelated and were key in distinguishing patient groups. Notably, all five compounds were integral to tryptophan metabolism, highlighting their shared biochemical pathway and significance in metabolic profiling. This is consistent with previous evidence reporting the presence of L-Tryptophan and its derivatives, such as indole compounds, in metabolic profiling related to diabetes 37 . One study found that tryptophan metabolism is critical in regenerating pancreatic β-cells, essential for insulin production in patients with T2DM 38 . Other studies supported the involvement of indole-3-carbinol in controlling the main components of metabolic dysfunction in T2DM, including insulin sensitivity and inflammation 39 – 41 . Consequently, the interconnection of these metabolites within the tryptophan metabolic pathway may suggest that intrinsic changes could be central to the metabolic alterations observed in diabetic patients​ 42 , 43 . Different metabolic markers were also more directly associated with diabetes progression. As previously discussed, BCAAs like valine and leucine, rather than tryptophan, have been associated with insulin resistance in some patient cohorts 44 , 45 . This emphasizes that while tryptophan metabolism is crucial, it may not be the main or primary pathway influenced in all cases of T2DM 46 ​. Interestingly, the study also revealed increased Pentose Phosphate Pathway (PPP) activation, particularly for Pyridoxal 5'-phosphate. This pathway mainly contributes to cellular redox balance and nucleotide synthesis by producing Nicotinamide Adenine Dinucleotide Phosphate (NADPH) and ribose-5-phosphate 47 . This suggests an adaptive response to oxidative stress commonly observed in T2DM. The suppression of metabolites obtained from L-Tryptophan metabolism further confirms the potential changes that could occur in this pathway in diabetic conditions. High levels of citric acid and cis-Aconitic acid, which are considered key intermediates of the TCA cycle, were also shown in this study. In contrast, Zou et al. reported low levels of these metabolites in individuals with T2DM 48 . Overall, this confirms the significant metabolic shifts that could occur in diabetic patients, suggesting a complex interaction between various metabolic pathways. Notably, our study reported a higher influence of vitamin B6 metabolism in the diabetic group. This confirms existing evidence reporting that deficiency in vitamin B6 is associated with increased insulin resistance and impaired glucose tolerance 49 . Lastly, the differential profiling of the metabolic expression between pre-diabetic and diabetic stage determines a positive correlation that differentiated between prediabetic/controlled diabetic and uncontrolled diabetic patients based on variation in 5-Dodecenoic acid, 4-Ethylbenzoic acid, Succinylacetone, Hexadecanedioic acid, 2-Phenylbutyric acid, Homovanillic acid, and phosphatidylcholine PC(18:1(9Z)/18:1(9Z)) expression levels between both groups. For example, we found that 5-Dodecenoic acid was elevated in uncontrolled diabetes, consistent with studies reporting fatty acid accumulation in poorly regulated diabetes due to impaired lipid oxidation and mitochondrial function 50 , 51 . In contrast, PC(18:1(9Z)/18:1(9Z)) was more prominent in controlled diabetes, supporting previous findings that connect stable phosphatidylcholine levels with better blood sugar control and balanced lipid levels 52 , 53 . This further confirms that metabolic regulation might be preserved in prediabetic and controlled diabetic states but is disrupted under uncontrolled conditions. Moreover, existing evidence reported the association between the elevated levels of Succinylacetone and Hexadecanedioic acid to mitochondrial dysfunction and oxidative stress observed in uncontrolled diabetes 54 . Overall, our findings emphasize that these specific metabolites, when analyzed together, can serve as robust markers for differentiating uncontrolled from controlled diabetes. Our study also indicated a strong correlation between Homovanillic acid and other metabolites, aligning with research highlighting its association with increased oxidative stress and inflammation in advanced diabetes 55 . Future research could evaluate these markers as therapeutic targets for improving metabolic control. Despite the strengths of the study, some limitations should be acknowledged. The relatively small sample size, particularly for the control group, may limit the statistical power and the generalizability of obtained findings. The geographically and ethnically specific study population could also limit the applicability of results to more diverse populations. Further validation studies with larger cohorts would be necessary to confirm the identified biomarkers and their potential clinical utility. Conclusion The application of advanced ML methodologies in this study highlights key metabolic markers and pathways that differentiate controlled from uncontrolled T2DM. Intrinsic metabolites, such as alpha-ketoisovaleric acid and uridine, were found to be linked to metabolic stress and inflammation, in addition to their correlation with tryptophan metabolism. Differential expressions of metabolites such as 5-Dodecenoic acid and PC (18:1(9Z)/18:1(9Z)) may also serve as markers for distinguishing between controlled and uncontrolled stages of diabetes. Consequently, the study’s findings offer novel insights into the application of ML methods to improve diabetes diagnosis and progression and identify potential biomarkers for individualized therapeutic intervention. Declarations Conflict of Interest : The authors declare no competing interests. Ethics Approval : This project was approved by the University Sharjah Hospital IRB committee. Approval number, HERC-012-10,062,019. Author Contribution Conception and design: M.T.A., A.S.B., J.F., and B.A.; Methodology: M.T.A., A.S.B, and J.F.; Acquisition of data: M.T.A., and N.A.; Software: A.S.B., and J.F.; Validation: M.T.A., and B.A.; Formal analysis, M.T.A., A.S.B., and J.F.; Investigation: M.T.A., A.S.B., and J.F.; Resources, M.T.A., and B.A.; Data curation, M.T.A., A.S.B., N.A., and J.F.; Drafting of the manuscript: M.T.A., A.S.B., J.F., A.M., and N.A. Critical revision of the manuscript: M.T.A., J.F., and B.A.; Statistical analysis: M.T.A., A.S.B., and J.F.; Administrative, technical, or logistic support: M.T.A., N.A. and B.A.; Supervision: M.T.A., N.A and B.A.; Others: All authors have read and agreed to the published version of the manuscript. Acknowledgement The authors would like to thank Khalifa University and University Sharjah Hospital for its support of this project. We also are grateful to all the study participants; without their contribution this work would not have been possible. Data Availability All data are presented in the paper References Zhao, Y. et al., The Prevalence of Metabolic Disease Multimorbidity and Its Associations With Spending and Health Outcomes in Middle-Aged and Elderly Chinese Adults. Frontiers Public. Health Volume 9–2021 (2021). Magliano Dj Fau - & Boyko E.J. & Boyko, E.J. Lin, X. et al., Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025. Saeedi, P. et al., Mortality attributable to diabetes in 20–79 years old adults, 2019 estimates: Results from the International Diabetes Federation Diabetes Atlas, 9(th) edition. Bellary, S. A. O., Kyrou, I. A. O., Brown, J. E. & Bailey, C. J. Type 2 diabetes mellitus in older adults: clinical considerations and management. Shahisavandi, M., Wang, K. A. O., Ghanbari, M. A. O. & Ahmadizar, F. A. O. Exploring Metabolomic Patterns in Type 2 Diabetes Mellitus and Response to Glucose-Lowering Medications-Review. LID – 10.3390/genes14071464 [doi] LID – 1464. DeForest, N. & Majithia, A. R. Genetics of Type 2 Diabetes: Implications from Large-Scale Studies. Ortiz-Martínez, M. et al., Recent Developments in Biomarkers for Diagnosis and Screening of Type 2 Diabetes Mellitus. Herder, C. et al., Immunological and cardiometabolic risk factors in the prediction of type 2 diabetes and coronary events: MONICA/KORA Augsburg case-cohort study. . Floegel, A. et al., Identification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach. Würtz, P. et al., Circulating metabolite predictors of glycemia in middle-aged men and women. Jin, Q. A. O. & Ma, R. A. .X. Metabolomics in Diabetes and Diabetic Complications: Insights from Epidemiological Studies. LID – 10.3390/cells10112832 [doi] LID – 2832. Arneth, B., Arneth, R. & Shams, M. Metabolomics of Type 1 and Type 2 Diabetes. LID – 10.3390/ijms20102467 [doi] LID – 2467. Ighodaro, O. M. Molecular pathways associated with oxidative stress in diabetes mellitus. Ding, L., Fan, L., Xu, X., Fu, J. & Xue, Y. Identification of core genes and pathways in type 2 diabetes mellitus by bioinformatics analysis. El-Kebbi, I. M., Bidikian, N. H., Hneiny, L. & Nasrallah, M. P. Epidemiology of type 2 diabetes in the Middle East and North Africa: Challenges and call for action. Kreienkamp, R. J., Voight, B. F., Gloyn, A. L. & Udler, M. S. Genetics of Type 2 Diabetes. BTI - Diabetes in America. Dahmani, K. M. A., Suliman, M., Hafidh, K., Beshyah, S. A. & Epidemiology Technology, and Professional Perspectives on Diabetes in the United Arab Emirate: A Focused Review. Journal Diabetes Endocr. Practice (2024). Banimfreg, B. H., Shamayleh, A., Alshraideh, H., Semreen, M. H. & Soares, N. C. Untargeted approach to investigating the metabolomics profile of type 2 diabetes emiratis. Schneider, A., Hommel, G., Fau - Blettner, M. & Blettner, M. Linear regression analysis: part 14 of a series on evaluation of scientific publications. Fox, J. Regression Diagnostics: An Introduction (SAGE, 2019). Biau, G. & Scornet, E. A random forest guided tour. TEST 25 , 197–227 (2016). Schonlau, M. & Zou, R. Y. The random forest algorithm for statistical learning. Stata J. 20 , 3–29 (2020). Xia, J. & Wishart, D. S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst. Wiener, A. L. .a.M. Classification and Regression by randomForest. R News . 2 , 18–22 (2002). Hong, H., Xiaoling, G. & Hua, Y. Variable selection using Mean Decrease Accuracy and Mean Decrease Gini based on Random Forest. in 7th IEEE International Conference on Software Engineering and Service Science (ICSESS) 219–224 (2016). 219–224 (2016). (2016). Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Ogata, H. et al., KEGG: Kyoto Encyclopedia of Genes and Genomes. . Yashi, K. & sharon, d. Obesity and Type 2 Diabetes. (2023). Klein, S., Gastaldelli, A., Yki-Järvinen, H. & Scherer, P. E. Why does obesity cause diabetes?. Manoli, I. & Venditti, C. P. Disorders of branched chain amino acid metabolism. Lynch, C. J. & Adams, S. H. Branched-chain amino acids in metabolic signalling and insulin resistance. Vanweert, F. A. O., Schrauwen, P. & Phielix, E. A. O. Role of branched-chain amino acid metabolism in the pathogenesis of obesity and type 2 diabetes-related metabolic disturbances BCAA metabolism in type 2 diabetes. Zhang, Y., Guo, S., Xie, C. & Fang, J. A. O. Uridine Metabolism and Its Role in Glucose, Lipid, and Amino Acid Homeostasis. Urasaki, Y., Pizzorno, G. & Le, T. T. Uridine affects liver protein glycosylation, insulin signaling, and heme biosynthesis. Yang, Y. A. O., Yu, T., Niu, Z. & Gao, L. A. O. The predictive value of plasma uridine for type 2 diabetes and its atherosclerotic complications. LID – 10.1530/EC-24-0075 [doi] LID - e240075. Qi, Q. A. O. et al., Host and gut microbial tryptophan metabolism and type 2 diabetes: an integrative analysis of host genetics, diet, gut microbiome and circulating metabolites in cohort studies. . Lee, Y. S., Lee, C., Choung, J. S., Jung, H. S. & Jun, H. A. -O. Glucagon-Like Peptide 1 Increases β-Cell Regeneration by Promoting α- to β-Cell Transdifferentiation. Hu, W. et al., Update of Indoles: Promising molecules for ameliorating metabolic diseases. . Centofanti, F. et al., Synthetic Methodologies and Therapeutic Potential of Indole-3-Carbinol (I3C) and Its Derivatives. in Pharmaceuticals, Vol. 16 (2023). Wang, Y. et al., Metabolites in the association between early-life famine exposure and type 2 diabetes in adulthood over a 5-year follow-up period. LID – 10.1136/bmjdrc-2020-001935 [doi] LID - e001935. . Lu, W. et al., Discovery of metabolic biomarkers for gestational diabetes mellitus in a Chinese population. Shatova, O. P., Shestopalov, A. V. & Tryptophan Metabolism A New Look at the Role of Tryptophan Derivatives in the Human Body. USB 143, 3–15 (2023). Arany, Z. & Neinast, M. Branched Chain Amino Acids in Metabolic Disease. De Bandt, J. A. O., Coumoul, X. A. O. & Barouki, R. Branched-Chain Amino Acids and Insulin Resistance, from Protein Supply to Diet-Induced Obesity. LID – 10.3390/nu15010068 [doi] LID – 68. Jiang, Y. et al., Gut Microbiota-Tryptophan Metabolism-GLP-1 Axis Participates in β-Cell Regeneration Induced by Dapagliflozin. . Moon, D. O. N. A. D. P. H. & Dynamics Linking Insulin Resistance and β-Cells Ferroptosis in Diabetes Mellitus. LID – 10.3390/ijms25010342 [doi] LID – 342. Zou, K. A. O. et al., Impaired glucose partitioning in primary myotubes from severely obese women with type 2 diabetes. . Zhu, J. et al. Folate, Vitamin B6, and Vitamin B12 Status in Association With Metabolic Syndrome Incidence. Cummins, M. A. O., Delmonte, G., Wechsler, S. & Schlesinger, J. J. Alleviating mitochondrial dysfunction in diabetic cardiomyopathy through the Adipsin and Irak2 pathways. Zhou, Y. et al., Metrnl Alleviates Lipid Accumulation by Modulating Mitochondrial Homeostasis in Diabetic Nephropathy. . Cai, Y., Qi, X., Zheng, Y., Zhang, J. & Su, H. Lipid profile alterations and biomarker identification in type 1 diabetes mellitus patients under glycemic control. Fu, Y. et al., Cordycepin Ameliorates High Fat Diet-Induced Obesity by Modulating Endogenous Metabolism and Gut Microbiota Dysbiosis. LID – 10.3390/nu16172859 [doi] LID – 2859. . Monnerie, S. et al., Metabolomic and Lipidomic Signatures of Metabolic Syndrome and its Physiological Components in Adults: A Systematic Review. . Liu, L. et al., Metabolic Homeostasis of Amino Acids and Diabetic Kidney Disease. LID – 10.3390/nu15010184 [doi] LID – 184. . Additional Declarations No competing interests reported. Supplementary Files SupplementaryTable1.pdf SupplementaryFigure2.tiff SupplementaryFigure1.tiff Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6606669","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":506682921,"identity":"49fb4e57-4ba1-4db0-bb8c-371f47a9bc10","order_by":0,"name":"Mohammad T. Al Bataineh","email":"","orcid":"","institution":"Yarmouk University","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"T. Al","lastName":"Bataineh","suffix":""},{"id":506682924,"identity":"4f4b6adf-800c-43d0-b77c-99c65a3a459f","order_by":1,"name":"Andre Sanches Barreiros","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Andre","middleName":"Sanches","lastName":"Barreiros","suffix":""},{"id":506682925,"identity":"09477d37-b778-4688-94cb-e53e9c578810","order_by":2,"name":"Joviana Farhat","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Joviana","middleName":"","lastName":"Farhat","suffix":""},{"id":506682930,"identity":"024058af-431a-41a3-a950-46aa5b94567a","order_by":3,"name":"Amal Mayyas","email":"","orcid":"","institution":"American University of Madaba","correspondingAuthor":false,"prefix":"","firstName":"Amal","middleName":"","lastName":"Mayyas","suffix":""},{"id":506682932,"identity":"d12835f8-1a52-4410-9892-d5f873261d94","order_by":4,"name":"Noura AlKhayyal","email":"","orcid":"","institution":"University Hospital of Sharjah","correspondingAuthor":false,"prefix":"","firstName":"Noura","middleName":"","lastName":"AlKhayyal","suffix":""},{"id":506682933,"identity":"872ecb0a-b0dd-4a20-964b-a9414ad10044","order_by":5,"name":"Basem Al-Omari","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+UlEQVRIiWNgGAWjYBACeygtAyYTKhIIazFsgNA8EC1niNBicABZC2MbMba0H3/46AaDHQ+/2OFnHx7OS0tsYD/8gOFHBW4t9jw5xsY5DMk8krPTjGckbstJbOBJM2DsOYPHloYcNukcBmYeg9sJxgyJ2yoSGxhyGBh42/D45fzz579zGOqBWtI/MyTOAWrhf8PA+PcfHi03EsyYcxgOA7XkAG1pADpMAmgpbwMeh814YyydY3Ac6JecYoaEY2nGbRLPDA7LHMPjff70h59zKqrl+KXTNzP+qEmW7edPfvjwTQ1uLVDnIbHZgPgAIQ2jYBSMglEwCvADABJHTHpWP6/YAAAAAElFTkSuQmCC","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Basem","middleName":"","lastName":"Al-Omari","suffix":""}],"badges":[],"createdAt":"2025-05-06 23:53:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6606669/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6606669/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90584432,"identity":"7b37776e-cc05-4843-a980-869c77712bdd","added_by":"auto","created_at":"2025-09-04 11:05:04","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":734495,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA:\u003c/strong\u003eGraph of Alpha-ketoisovaleric acid significance in distinguishing between patients diagnosed vs undiagnosed with T2DM.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB:\u003c/strong\u003eTable of Alpha-ketoisovaleric acid significance in distinguishing between patients diagnosed vs undiagnosed with T2DM.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/566023a4e1c16443955a7ed8.jpg"},{"id":90585515,"identity":"c14a632a-59b3-4f53-a0ca-937766c02559","added_by":"auto","created_at":"2025-09-04 11:21:04","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":480331,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e: Pathway enrichment analysis of the most significant metabolic pathways associated with the significant metabolites in humans.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB: \u003c/strong\u003eTable of the most significant metabolic pathways associated with the significant metabolites in humans.\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/4848c34164d9a1db011e1149.jpg"},{"id":90584436,"identity":"165d5d18-f9e4-4101-b5e1-bc251e89a691","added_by":"auto","created_at":"2025-09-04 11:05:04","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":625520,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA:\u003c/strong\u003e The metabolic superpathway for the metabolites given in Figure 2B and Supplementary Table 1 along with their associated pathways are depicted in Figure 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB: \u003c/strong\u003eTable of the significant metabolites from panel A, which is a subset of those shown in Figure 1 and Supplementary Table 1.\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/5af91f260ff3f7e3dcdf3220.jpg"},{"id":103228915,"identity":"6013c758-22cc-4f69-9830-1b58e8bcf83e","added_by":"auto","created_at":"2026-02-23 11:41:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2525353,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/438675b9-e3e8-4198-902f-5ef215ada929.pdf"},{"id":90584429,"identity":"d0b31419-cfcb-467d-9080-01c8555ea3f9","added_by":"auto","created_at":"2025-09-04 11:05:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":90088,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/01cc34e2b88a208ae3e50459.pdf"},{"id":90585135,"identity":"8a3c69be-ad59-47a5-aa04-f29de21c5ba1","added_by":"auto","created_at":"2025-09-04 11:13:04","extension":"tiff","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":867048,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/345ed7c6376a43bceb9c9fb4.tiff"},{"id":90584434,"identity":"c56c3d20-959f-4516-b37b-9dcc6ea2f591","added_by":"auto","created_at":"2025-09-04 11:05:04","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1017686,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-6606669/v1/2793f397bda8af5a2f732581.tiff"}],"financialInterests":"No competing interests reported.","formattedTitle":"Metabolic Fingerprints of Diabetes: Machine Learning Reveals Distinct Biomarkers in Type 2 Diabetes Mellitus","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMetabolic disorders have been a major public health issue and a clinical challenge worldwide \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Type 2 Diabetes Mellitus (T2DM) is one of the most common metabolic diseases that affects approximately 422\u0026ndash;537\u0026nbsp;million people worldwide \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. In 2021, it was estimated that approximately 6.7\u0026nbsp;million deaths were caused by T2DM and its complications \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, making it one of the top ten causes of mortality globally \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. At a regional level, the Middle East and North Africa region (MENA) had a high prevalence of DM in 2019 (12.2%), which is expected to increase by 96% between 2019 and 2045 \u003csup\u003e4\u003c/sup\u003e. In clinical practice, T2DM is the most diagnosed type of diabetes, accounting for approximately 90% of all cases \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. T2DM is characterized by a gained physiological insulin resistance due to islet cell dysfunction \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe persistent increase in the burden of T2DM and its complications highlights the need for a deeper understanding of the underlying metabolic disturbances. Therefore, identifying biomarkers during the screening and prediction stages can allow for more personalized healthcare management \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. In practice, multiple biomarkers, including fasting plasma glucose (FPG), glycated hemoglobin A1c (HbA1c), triglycerides, HDL cholesterol, inflammatory mediators, adiponectin, liver enzymes, and fetuin-A have been used for estimating T2DM risk \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. However, many of these biomarkers do not fully capture the complexity of T2DM etiology \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Therefore, metabolomics has been applied to identify and quantify metabolites as they can better reflect physiological dysfunctions and allow the earlier detection of T2DM \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Metabolomics studies have revealed several blood sugars, sugar-related metabolites, glycolysis/gluconeogenesis pathway components, and tricarboxylic acid (TCA) cycle intermediates associated with diabetes \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Other studies revealed novel biomarkers associated with T2DM, such as blood acylcarnitines, phenylalanine metabolites, energy metabolism, and lipid metabolism \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Branched-chain amino acids (BCAA) such as valine, leucine, isoleucine, and phospholipids were among the most popular biomarkers associated with T2DM progression \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eSeveral molecular pathways are also involved in the pathophysiology of T2DM through the induction of oxidative stress \u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Accordingly, alterations in these signaling pathways have been linked to disease progression \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. The limited understanding of the specific molecular mechanisms and epidemiological status underlying T2DM in the MENA region may hinder effective treatment strategies \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Both genetic and environmental factors play an important role in determining an individual's risk of developing T2DM \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. Identifying genetic variations associated with increased disease risk could be essential for improving disease management, enhancing clinical outcomes, and preventing complications \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn practice, evaluating the intrinsic pathways during the course of T2DM is needed for non-invasive diagnosis and long-term targeted therapy \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. This is particularly relevant to patients in the United Arab Emirates (UAE) who are at high risk of T2DM and its complications \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Thus, our study aims to analyze preliminary data from Emirati diabetic patients with uncontrolled and controlled disease status using more advanced methodologies. Specifically, it employs univariate analysis of covariant alongside the powerful machine-learning AI tool, Random Forest (RF). As a regression tree technique, RF uses bootstrap aggregation and predictor randomization to achieve high predictive accuracy. This approach is particularly effective in handling complex data-generating processes to ensure optimal model performance.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy Participants\u003c/h2\u003e\u003cp\u003eA case-control study was conducted on adult patients recruited from the Diabetes and Endocrine Center at the University Hospital Sharjah in the UAE. The cohort was defined as patients clinically diagnosed with T2DM and with a Hemoglobin Subunit Alpha 1 (HbA1)\u0026thinsp;\u0026ge;\u0026thinsp;5.7%. Participants with an HbA1c\u0026thinsp;\u0026ge;\u0026thinsp;5.7% but were not clinically confirmed to be diabetic were chosen as controls. The approval for the study conduction was obtained from the hospital\u0026rsquo;s Research and Ethics Committee. All experiments were performed in accordance with relevant guidelines and regulations.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eSample collection, preparation, and analysis\u003c/h3\u003e\n\u003cp\u003eBlood specimens were carefully procured from the patients, each of whom had given informed consent and underwent the sampling procedure in a controlled clinical setting under the supervision of trained medical personnel. In the current study, the protocols for sample preparation and analysis adhered to the methodologies previously employed in a similar study \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eThe patient's data was reanalyzed using statistical tools on R version 4.2.2. The metabolic concentration values were normalized as shown in MetaboAnalyst\u0026reg; 5.0, where data was first cleaned, then standardized and normalized through the logarithmic transformation. To address the age bias between the cohorts, a conditional linear regression was employed based on the provided data \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. The regression was performed under the condition that the following conditions were met: Primarily, the gradient of the best-fit line or the control patients (m\u003csub\u003ec\u003c/sub\u003e) had to be within 2 standard deviations from the gradient of the best-fit line for all patients (m\u003csub\u003ea\u003c/sub\u003e). Subsequently, the mean metabolite value for control and case patients had to be at least 2 standard deviations apart. Finally, the correlation coefficient for the control patients is greater than |0.3|. These conditions exist to ensure that only the metabolites that show a linear correlation with age would be corrected. Otherwise, some correction artifacts may arise due to the strong correlation between the two variables, age and T2DM.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eRandom Forest (RF) in machine learning (ML)\u003c/h3\u003e\n\u003cp\u003eThe RF machine learning algorithm from the R library was implemented afterward to identify a ranked list of metabolites contributing to the separation between both groups. RF is a type of general-purpose classification and regression ML method that can be performed in settings where the number of variables exceeds the number of observables \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. The algorithm combines several randomized decision trees in the learning phase and averages their predictions to obtain a final result\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe RF method applied a pseudo-random division of the patients\u0026rsquo; data set into training and test sets. A seed ranging from 1 to 100 (all-inclusive) split the data into two sets. The training set comprised 80% of the patients and was used to \u0026lsquo;teach\u0026rsquo; the algorithm to distinguish between the different groups based on their 148 metabolites. The RF algorithm is then tested on the test set (20% of the data) by attempting to categorize the patients based on their metabolites correctly. Each run weighs the metabolite effects differently depending on the training data set. Therefore, it is necessary to average the resulting metabolites throughout the runs. This was done by averaging the Mean Decrease Gini for each metabolite across all the runs, from which an ordered list of the most significant metabolites can be generated.\u003c/p\u003e\n\u003ch3\u003eCovariance analytical method\u003c/h3\u003e\n\u003cp\u003eIn addition to the RF method described above, a covariance analysis was performed to further quantify the significant metabolites distinguishing between the two patient cohorts. A one-way ANOVA test compared the normalized metabolite concentration between the controls and the cases. A metabolite was considered significantly correlated if the linear correlation between its concentrations and the patient cohort had a False Discovery Rate (FDR) p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05. This method quantifies the linear dependence for each metabolite on the cohorts, and the strength of this correlation is recorded.\u003c/p\u003e\u003cp\u003eA secondary aspect investigated with the linear covariance method was the correlation between different metabolites. Strong correlations (|R| \u0026gt;0.8) would indicate metabolites that can be grouped together, suggesting involvement in similar metabolic pathways.\u003c/p\u003e\u003cp\u003eA tertiary aspect of this method was identifying metabolites that strongly correlated with patients having 5.7% \u0026le; HbA1c\u0026thinsp;\u0026le;\u0026thinsp;6.4% and HbA1c\u0026thinsp;\u0026ge;\u0026thinsp;6.5%. These metabolites could help distinguish between prediabetic/controlled and uncontrolled diabetics.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003ePathway analysis\u003c/h2\u003e\u003cp\u003eA combined list of significant metabolites from the RF and the covariant methods was created, followed by pathway analysis. Pathway analysis was performed using the online tool MetaboAnalyst\u0026reg; 5.0 \u003csup\u003e24\u003c/sup\u003e. The analysis was done with the following settings: the enrichment method was set to a scatter plot visualization method, the Hypergeometric test was applied, the topological analysis used relative betweenness Centrality, and the Homo sapiens Kyoto Encyclopedia of Genes and Genome (KEGG) pathway library was used. The resulting figure displays the pathways in which the metabolites are involved.\u003c/p\u003e\u003cp\u003eThe final step in this study was to combine the pathways from the pathway analysis into a super pathway. This was done by linking the most significant pathways from the analysis through the key metabolites identified by the RF and covariant methods, in conjunction with the KEGG library. A table listing the metabolites was provided, showing the fold Log2 change and the FDR q-value of the metabolite. A positive fold Log2 indicates an amplified metabolite in the cases cohort or a suppressed metabolite in the control cohort, while a negative fold Log2 indicates the opposite.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003eStudy participants\u003c/h2\u003e\u003cp\u003eSeventy-six subjects were voluntarily enrolled in this study; 50 were diagnosed with T2DM and treated at the University Hospital Sharjah. None of the other 26 patients had been diagnosed with T2DM and were considered the control group in this study. The evaluation of study participants resulted in significant differences between case and control groups that are relevant to detecting T2DM biomarkers. Most cases were older than 65 years, while most controls were under 65 years, indicating a strong association between age and T2DM progression (p\u0026thinsp;=\u0026thinsp;4.10 x 10⁻⁸). Additionally, a higher proportion of cases had a BMI\u0026thinsp;\u0026gt;\u0026thinsp;25 kg/m\u0026sup2; compared to controls, yielding a borderline significant difference in BMI between both groups (p\u0026thinsp;=\u0026thinsp;0.050). HbA1c levels were also significantly different between the groups (p\u0026thinsp;=\u0026thinsp;0.024), with more cases having HbA1c levels\u0026thinsp;\u0026ge;\u0026thinsp;6.5%, while a higher percentage of controls fell within the pre-diabetic range (5.7% \u0026le; HbA1c\u0026thinsp;\u0026le;\u0026thinsp;6.4%) (See Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe characteristics of study participants\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eVariables\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCases (n\u0026thinsp;=\u0026thinsp;50)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eControls (n\u0026thinsp;=\u0026thinsp;26)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ep-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003en (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003en (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003e0.945\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e15 (30.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (30.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e35 (70.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e18 (69.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAge\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003e4.10 x 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u0026thinsp;\u0026lt;\u0026thinsp;65 years\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e13 (26.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24 (92.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u0026thinsp;\u0026ge;\u0026thinsp;65 years\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e37 (74.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2 (7.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eBMI\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003e0.050\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u0026thinsp;\u0026le;\u0026thinsp;25 kg/m2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e8 (16.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (30.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u0026thinsp;\u0026gt;\u0026thinsp;25 kg/m2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e34 (68.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e18 (69.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eN/A (Wheelchair)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e8 (16.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eBlood sugar level (HbA1c)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e5.7% \u0026le; HbA1c\u0026thinsp;\u0026le;\u0026thinsp;6.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e23 (46.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19 (73.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.024\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHbA1c\u0026thinsp;\u0026ge;\u0026thinsp;6.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e27 (54.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7 (26.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eRandom Forest (RF) in machine learning (ML)\u003c/h2\u003e\u003cp\u003eThe first method used to identify the metabolites that distinguish the two groups (diagnosed vs undiagnosed T2DM) is the RF algorithm from the R library \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. This method ranks metabolites based on their importance in categorizing patients, producing a list from the most to the least significant. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003eA \u0026amp; B shows an example of this for a particular run.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003eA \u0026amp; B shows that for seed 100, the RF method identifies Alpha-ketoisovaleric acid, 2-pyrrolidinone, and Uridine as the most significant metabolites, respectively. However, different runs may rank the metabolites differently, as the order highly depends on how the seed splits the training and testing set. The RF method can correctly classify patients as not diagnosed vs undiagnosed with T2DM, with an average accuracy of 88.0% across 1000 runs.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eCovariance analytical method\u003c/h2\u003e\u003cp\u003eAlpha-ketoisovaleric acid is the most significant metabolite for the covariance method.\u003c/p\u003e\u003cp\u003eIn addition to the RF method, the covariance (using one-way ANOVA with FDR q-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05) was applied to assess the relationship between metabolites and diagnosis. The covariance method complements the RF method by determining correlations and providing p-values for each metabolite. The results are shown in Supplementary Table\u0026nbsp;1.\u003c/p\u003e\u003cp\u003eSupplementary Table\u0026nbsp;1 lists the metabolites Alpha-ketoisovaleric acid, N-Methylhydantoin, 3-Methylindole, Indole-3-carbinol and L-Tryptophan as the most significant for the one-way ANOVA test. This result aligns with Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e, where these same metabolites were also identified as the most significant by the RF method.\u003c/p\u003e\u003cp\u003eAlpha-ketoisovaleric acid, N-Methylhydantoin, Indole-3-carbinol, L-Tryptophan, and Indole are all strongly (positively) correlated with each other, with R\u0026thinsp;\u0026gt;\u0026thinsp;0.9 (Supplementary Fig.\u0026nbsp;1). Together, these five metabolites are the most significant in identifying patient groupings. These five metabolites (except for Alpha-ketoisovaleric acid) and Indoleacetic acid, 5-Methoxytryptophol, and Indolelactic acid are all involved in the Tryptophan metabolism \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. However, unlike metabolites like N-Methylhydantoin, these metabolites are overexpressed in the cases of patients.\u003c/p\u003e\u003cp\u003eThe metabolites 5-Dodecenoic acid, 4-Ethylbenzoic acid, Succinylacetone, Hexadecanedioic acid, 2-Phenylbutyric acid, Homovanillic acid, and PC (18:1(9Z)/18:1(9Z)) distinguish prediabetic/controlled diabetic from uncontrolled diabetic patients, as shown in Supplementary Fig.\u0026nbsp;2. These metabolites are also strongly (positively) correlated (\u0026gt;\u0026thinsp;0.8) with each other and may be influenced by the blood sugar level bias observed in the data in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003ePathway analysis\u003c/h2\u003e\u003cp\u003eThe RF and covariance methods have identified significant metabolites distinguishing between diagnosed and undiagnosed diabetic patients. The next step is to identify the most significant metabolic pathways involving these metabolites through the pathway analysis method in MetaboAnalyst 5.0. All metabolites listed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Supplementary Table\u0026nbsp;1 were processed using the KEGG Homo sapiens pathway library \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e to find the significant metabolic pathways. The resulting pathways are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, the most impactful pathways with a fold change\u0026thinsp;\u0026gt;\u0026thinsp;1 include Caffeine metabolism, Glyoxylate and Dicarboxylate metabolism, Aminoacyl-tRNA biosynthesis, TCA cycle, Linoleic acid metabolism, Tryptophan metabolism, Valine, leucine and isoleucine biosynthesis, and Vitamin B6 metabolism. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, the pathways, Caffeine metabolism, and valine, leucine, and isoleucine, biosynthesis have a p-value\u0026thinsp;\u0026gt;\u0026thinsp;0.05 and thus are not considered significant. The table also shows that Linoleic acid metabolism and Aminoacyl-tRNA biosynthesis have zero impact and are insignificant. The remaining pathways were considered when constructing the combined superpathway.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eSuperpathway construction\u003c/h2\u003e\u003cp\u003eUsing the most significant pathways from Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e and metabolites from Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Supplementary Table\u0026nbsp;1, a link between these can be established into one superpathway to provide a possible metabolic picture of the biological processes at work. This is represented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e3\u003c/span\u003eA \u0026amp; B as a network, with all connections made using the KEGG library \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe superpathway created shows that the significant pathways from Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e are connected through the Glycolysis pathway. It begins with Glucose (which is not a metabolite in the human body) and converts it into Alpha-ketoisovaleric acid after several intermediate metabolites.\u003c/p\u003e\u003cp\u003eSome of these intermediate metabolites branch off into Pyridoxal 5\u0026rsquo;-phosphate via the Vitamin B6 metabolism pathway. Later intermediate metabolites can be converted into Glycine or L-Tryptophan instead, continuing down the Tryptophan metabolism pathway. As shown on the Table in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e3\u003c/span\u003eB, the end products of this pathway, including Indole-3-carbinol, Indole, and 3-methylindole, have a negative fold-change, indicating a higher concertation in the control cohort. Conversely, the pathway may lead to metabolites such as 5-methoxytryptophol, N-Acetylserotonin, Indolelactic acid, and Indoleacetic acid, with a positive fold-change, indicating higher concentrations in the case cohort. The final metabolite of glycolysis, Alpha-ketoisovaleric acid, can then enter the TCA cycle via citric acid, which is converted into cis-aconitic acid. In the reverse direction, cis-aconitic acid can convert into citric acid, as part of the Glyoxylate and dicarboxylate metabolism pathway.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe investigated data from diabetic patients with controlled and uncontrolled T2DM utilizing sophisticated approaches, such as ML methods, bootstrap aggregation, and randomizing predictors to maximize predictive accuracy. The evaluation of study participants revealed differences in age, BMI, and HBA1c levels between cases and controls. Prior studies showed a strong association between higher BMI and increased risk of T2DM, in which obesity constitutes a critical factor for disease development \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. However, a more complex relationship between BMI and diabetes risk was previously reported, suggesting that factors such as muscle mass and distribution of fat may also play significant roles in disease progression \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. This highlights the importance of BMI and glycemic control as key factors in understanding the metabolic differences between cases and controls, reinforcing their relevance when assessing novel biomarkers in T2DM.\u003c/p\u003e\u003cp\u003eFirst, we explored the significantly expressed metabolic markers in diabetic subgroups. The study shows that alpha-ketoisovaleric acid, 2-Pyrrolidinone, and uridine are the most significantly expressed metabolic markers in patients with a confirmed T2DM diagnosis. Valine, Leucine, and Isoleucine biosynthesis were mainly upregulated in the diabetic group. This aligns with existing evidence reporting the association between elevated levels of alpha-ketoisovaleric acid and the progression of metabolic disorders, including diabetes \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Therefore, accumulating branched-chain amino acids (BCAA) intermediates such as alpha-ketoisovaleric acid may contribute to metabolic stress and inflammation that can exacerbate diabetes progression \u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. A recent study highlighted the link between increased circulating BCAA metabolites, including alpha-ketoisovaleric acid, and the risk of developing insulin resistance and T2DM \u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. It has also been suggested that uridine can influence insulin signaling \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, with potential implications for glucose homeostasis in T2DM. Urasaki et al. reported that uridine was shown to regulate glucose homeostasis by playing a dual role, as both high and low levels can affect insulin sensitivity \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e, presenting uridine as a potential modulator in metabolic diseases like T2DM. Recently, fasting uridine was found to be closely associated with carotid atherosclerosis in patients with T2DM, constituting a promising predictor of diabetes \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eNext, to explore the relevant correlation of metabolic pathways in diabetic subgroups, we determined that Alpha-ketoisovaleric acid, N-Methylhydantoin, Indole-3-carbinol, L-Tryptophan, and Indole exhibit strong positive correlations. These metabolites were highly interrelated and were key in distinguishing patient groups. Notably, all five compounds were integral to tryptophan metabolism, highlighting their shared biochemical pathway and significance in metabolic profiling. This is consistent with previous evidence reporting the presence of L-Tryptophan and its derivatives, such as indole compounds, in metabolic profiling related to diabetes \u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. One study found that tryptophan metabolism is critical in regenerating pancreatic β-cells, essential for insulin production in patients with T2DM \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Other studies supported the involvement of indole-3-carbinol in controlling the main components of metabolic dysfunction in T2DM, including insulin sensitivity and inflammation \u003csup\u003e\u003cspan additionalcitationids=\"CR40\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Consequently, the interconnection of these metabolites within the tryptophan metabolic pathway may suggest that intrinsic changes could be central to the metabolic alterations observed in diabetic patients​ \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Different metabolic markers were also more directly associated with diabetes progression. As previously discussed, BCAAs like valine and leucine, rather than tryptophan, have been associated with insulin resistance in some patient cohorts \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. This emphasizes that while tryptophan metabolism is crucial, it may not be the main or primary pathway influenced in all cases of T2DM \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e​. Interestingly, the study also revealed increased Pentose Phosphate Pathway (PPP) activation, particularly for Pyridoxal 5'-phosphate. This pathway mainly contributes to cellular redox balance and nucleotide synthesis by producing Nicotinamide Adenine Dinucleotide Phosphate (NADPH) and ribose-5-phosphate \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. This suggests an adaptive response to oxidative stress commonly observed in T2DM. The suppression of metabolites obtained from L-Tryptophan metabolism further confirms the potential changes that could occur in this pathway in diabetic conditions. High levels of citric acid and cis-Aconitic acid, which are considered key intermediates of the TCA cycle, were also shown in this study. In contrast, Zou et al. reported low levels of these metabolites in individuals with T2DM \u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. Overall, this confirms the significant metabolic shifts that could occur in diabetic patients, suggesting a complex interaction between various metabolic pathways. Notably, our study reported a higher influence of vitamin B6 metabolism in the diabetic group. This confirms existing evidence reporting that deficiency in vitamin B6 is associated with increased insulin resistance and impaired glucose tolerance \u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eLastly, the differential profiling of the metabolic expression between pre-diabetic and diabetic stage determines a positive correlation that differentiated between prediabetic/controlled diabetic and uncontrolled diabetic patients based on variation in 5-Dodecenoic acid, 4-Ethylbenzoic acid, Succinylacetone, Hexadecanedioic acid, 2-Phenylbutyric acid, Homovanillic acid, and phosphatidylcholine PC(18:1(9Z)/18:1(9Z)) expression levels between both groups. For example, we found that 5-Dodecenoic acid was elevated in uncontrolled diabetes, consistent with studies reporting fatty acid accumulation in poorly regulated diabetes due to impaired lipid oxidation and mitochondrial function \u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. In contrast, PC(18:1(9Z)/18:1(9Z)) was more prominent in controlled diabetes, supporting previous findings that connect stable phosphatidylcholine levels with better blood sugar control and balanced lipid levels \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. This further confirms that metabolic regulation might be preserved in prediabetic and controlled diabetic states but is disrupted under uncontrolled conditions. Moreover, existing evidence reported the association between the elevated levels of Succinylacetone and Hexadecanedioic acid to mitochondrial dysfunction and oxidative stress observed in uncontrolled diabetes \u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Overall, our findings emphasize that these specific metabolites, when analyzed together, can serve as robust markers for differentiating uncontrolled from controlled diabetes. Our study also indicated a strong correlation between Homovanillic acid and other metabolites, aligning with research highlighting its association with increased oxidative stress and inflammation in advanced diabetes \u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. Future research could evaluate these markers as therapeutic targets for improving metabolic control. Despite the strengths of the study, some limitations should be acknowledged. The relatively small sample size, particularly for the control group, may limit the statistical power and the generalizability of obtained findings. The geographically and ethnically specific study population could also limit the applicability of results to more diverse populations. Further validation studies with larger cohorts would be necessary to confirm the identified biomarkers and their potential clinical utility.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe application of advanced ML methodologies in this study highlights key metabolic markers and pathways that differentiate controlled from uncontrolled T2DM. Intrinsic metabolites, such as alpha-ketoisovaleric acid and uridine, were found to be linked to metabolic stress and inflammation, in addition to their correlation with tryptophan metabolism. Differential expressions of metabolites such as 5-Dodecenoic acid and PC (18:1(9Z)/18:1(9Z)) may also serve as markers for distinguishing between controlled and uncontrolled stages of diabetes. Consequently, the study\u0026rsquo;s findings offer novel insights into the application of ML methods to improve diabetes diagnosis and progression and identify potential biomarkers for individualized therapeutic intervention.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003e\u003cb\u003eConflict of Interest\u003c/b\u003e:\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003e\u003cb\u003eEthics Approval\u003c/b\u003e:\u003c/strong\u003e\u003cp\u003e This project was approved by the University Sharjah Hospital IRB committee. Approval number, HERC-012-10,062,019.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eConception and design: M.T.A., A.S.B., J.F., and B.A.; Methodology: M.T.A., A.S.B, and J.F.; Acquisition of data: M.T.A., and N.A.; Software: A.S.B., and J.F.; Validation: M.T.A., and B.A.; Formal analysis, M.T.A., A.S.B., and J.F.; Investigation: M.T.A., A.S.B., and J.F.; Resources, M.T.A., and B.A.; Data curation, M.T.A., A.S.B., N.A., and J.F.; Drafting of the manuscript: M.T.A., A.S.B., J.F., A.M., and N.A. Critical revision of the manuscript: M.T.A., J.F., and B.A.; Statistical analysis: M.T.A., A.S.B., and J.F.; Administrative, technical, or logistic support: M.T.A., N.A. and B.A.; Supervision: M.T.A., N.A and B.A.; Others: All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors would like to thank Khalifa University and University Sharjah Hospital for its support of this project. We also are grateful to all the study participants; without their contribution this work would not have been possible.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll data are presented in the paper\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eZhao, Y. et al., \u003cem\u003eThe Prevalence of Metabolic Disease Multimorbidity and Its Associations With Spending and Health Outcomes in Middle-Aged and Elderly Chinese Adults. Frontiers Public. Health Volume 9\u0026ndash;2021\u003c/em\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMagliano Dj Fau - \u0026amp; Boyko E.J. \u0026amp; Boyko, E.J.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLin, X. et al., \u003cem\u003eGlobal, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSaeedi, P. et al., \u003cem\u003eMortality attributable to diabetes in 20\u0026ndash;79 years old adults, 2019 estimates: Results from the International Diabetes Federation Diabetes Atlas, 9(th) edition.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBellary, S. A. O., Kyrou, I. A. O., Brown, J. E. \u0026amp; Bailey, C. J. Type 2 diabetes mellitus in older adults: clinical considerations and management.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShahisavandi, M., Wang, K. A. O., Ghanbari, M. A. O. \u0026amp; Ahmadizar, F. A. O. Exploring Metabolomic Patterns in Type 2 Diabetes Mellitus and Response to Glucose-Lowering Medications-Review. LID \u0026ndash;\u0026thinsp;10.3390/genes14071464 [doi] LID \u0026ndash;\u0026thinsp;1464.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDeForest, N. \u0026amp; Majithia, A. R. Genetics of Type 2 Diabetes: Implications from Large-Scale Studies.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOrtiz-Mart\u0026iacute;nez, M. et al., \u003cem\u003eRecent Developments in Biomarkers for Diagnosis and Screening of Type 2 Diabetes Mellitus.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHerder, C. et al., \u003cem\u003eImmunological and cardiometabolic risk factors in the prediction of type 2 diabetes and coronary events: MONICA/KORA Augsburg case-cohort study.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFloegel, A. et al., \u003cem\u003eIdentification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eW\u0026uuml;rtz, P. et al., \u003cem\u003eCirculating metabolite predictors of glycemia in middle-aged men and women.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJin, Q. A. O. \u0026amp; Ma, R. A. .X. Metabolomics in Diabetes and Diabetic Complications: Insights from Epidemiological Studies. LID \u0026ndash;\u0026thinsp;10.3390/cells10112832 [doi] LID \u0026ndash;\u0026thinsp;2832.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArneth, B., Arneth, R. \u0026amp; Shams, M. Metabolomics of Type 1 and Type 2 Diabetes. LID \u0026ndash;\u0026thinsp;10.3390/ijms20102467 [doi] LID \u0026ndash;\u0026thinsp;2467.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIghodaro, O. M. Molecular pathways associated with oxidative stress in diabetes mellitus.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDing, L., Fan, L., Xu, X., Fu, J. \u0026amp; Xue, Y. Identification of core genes and pathways in type 2 diabetes mellitus by bioinformatics analysis.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEl-Kebbi, I. M., Bidikian, N. H., Hneiny, L. \u0026amp; Nasrallah, M. P. Epidemiology of type 2 diabetes in the Middle East and North Africa: Challenges and call for action.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKreienkamp, R. J., Voight, B. F., Gloyn, A. L. \u0026amp; Udler, M. S. Genetics of Type 2 Diabetes. BTI - Diabetes in America.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDahmani, K. M. A., Suliman, M., Hafidh, K., Beshyah, S. A. \u0026amp; Epidemiology Technology, and Professional Perspectives on Diabetes in the United Arab Emirate: A Focused Review. \u003cem\u003eJournal Diabetes Endocr. Practice\u003c/em\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBanimfreg, B. H., Shamayleh, A., Alshraideh, H., Semreen, M. H. \u0026amp; Soares, N. C. Untargeted approach to investigating the metabolomics profile of type 2 diabetes emiratis.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchneider, A., Hommel, G., Fau - Blettner, M. \u0026amp; Blettner, M. Linear regression analysis: part 14 of a series on evaluation of scientific publications.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFox, J. \u003cem\u003eRegression Diagnostics: An Introduction\u003c/em\u003e (SAGE, 2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBiau, G. \u0026amp; Scornet, E. A random forest guided tour. \u003cem\u003eTEST\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 197\u0026ndash;227 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchonlau, M. \u0026amp; Zou, R. Y. The random forest algorithm for statistical learning. \u003cem\u003eStata J.\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e, 3\u0026ndash;29 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXia, J. \u0026amp; Wishart, D. S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWiener, A. L. .a.M. Classification and Regression by randomForest. \u003cem\u003eR News\u003c/em\u003e. \u003cb\u003e2\u003c/b\u003e, 18\u0026ndash;22 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHong, H., Xiaoling, G. \u0026amp; Hua, Y. Variable selection using Mean Decrease Accuracy and Mean Decrease Gini based on Random Forest. in \u003cem\u003e7th IEEE International Conference on Software Engineering and Service Science (ICSESS)\u003c/em\u003e 219\u0026ndash;224 (2016). 219\u0026ndash;224 (2016). (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKanehisa, M. \u0026amp; Goto, S. KEGG: kyoto encyclopedia of genes and genomes.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOgata, H. et al., \u003cem\u003eKEGG: Kyoto Encyclopedia of Genes and Genomes.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYashi, K. \u0026amp; sharon, d. Obesity and Type 2 Diabetes. (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKlein, S., Gastaldelli, A., Yki-J\u0026auml;rvinen, H. \u0026amp; Scherer, P. E. Why does obesity cause diabetes?.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eManoli, I. \u0026amp; Venditti, C. P. Disorders of branched chain amino acid metabolism.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLynch, C. J. \u0026amp; Adams, S. H. Branched-chain amino acids in metabolic signalling and insulin resistance.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVanweert, F. A. O., Schrauwen, P. \u0026amp; Phielix, E. A. O. Role of branched-chain amino acid metabolism in the pathogenesis of obesity and type 2 diabetes-related metabolic disturbances BCAA metabolism in type 2 diabetes.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, Y., Guo, S., Xie, C. \u0026amp; Fang, J. A. O. Uridine Metabolism and Its Role in Glucose, Lipid, and Amino Acid Homeostasis.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUrasaki, Y., Pizzorno, G. \u0026amp; Le, T. T. Uridine affects liver protein glycosylation, insulin signaling, and heme biosynthesis.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, Y. A. O., Yu, T., Niu, Z. \u0026amp; Gao, L. A. O. The predictive value of plasma uridine for type 2 diabetes and its atherosclerotic complications. LID \u0026ndash;\u0026thinsp;10.1530/EC-24-0075 [doi] LID - e240075.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQi, Q. A. O. et al., \u003cem\u003eHost and gut microbial tryptophan metabolism and type 2 diabetes: an integrative analysis of host genetics, diet, gut microbiome and circulating metabolites in cohort studies.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee, Y. S., Lee, C., Choung, J. S., Jung, H. S. \u0026amp; Jun, H. A. -O. Glucagon-Like Peptide 1 Increases β-Cell Regeneration by Promoting α- to β-Cell Transdifferentiation.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu, W. et al., \u003cem\u003eUpdate of Indoles: Promising molecules for ameliorating metabolic diseases.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCentofanti, F. et al., \u003cem\u003eSynthetic Methodologies and Therapeutic Potential of Indole-3-Carbinol (I3C) and Its Derivatives. in Pharmaceuticals, Vol. 16\u003c/em\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, Y. et al., \u003cem\u003eMetabolites in the association between early-life famine exposure and type 2 diabetes in adulthood over a 5-year follow-up period. LID \u0026ndash;\u0026thinsp;10.1136/bmjdrc-2020-001935 [doi] LID - e001935.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu, W. et al., \u003cem\u003eDiscovery of metabolic biomarkers for gestational diabetes mellitus in a Chinese population.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShatova, O. P., Shestopalov, A. V. \u0026amp; Tryptophan Metabolism A New Look at the Role of Tryptophan Derivatives in the Human Body. \u003cem\u003eUSB\u003c/em\u003e 143, 3\u0026ndash;15 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArany, Z. \u0026amp; Neinast, M. Branched Chain Amino Acids in Metabolic Disease.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDe Bandt, J. A. O., Coumoul, X. A. O. \u0026amp; Barouki, R. Branched-Chain Amino Acids and Insulin Resistance, from Protein Supply to Diet-Induced Obesity. LID \u0026ndash;\u0026thinsp;10.3390/nu15010068 [doi] LID \u0026ndash;\u0026thinsp;68.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiang, Y. et al., \u003cem\u003eGut Microbiota-Tryptophan Metabolism-GLP-1 Axis Participates in β-Cell Regeneration Induced by Dapagliflozin.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMoon, D. O. N. A. D. P. H. \u0026amp; Dynamics Linking Insulin Resistance and β-Cells Ferroptosis in Diabetes Mellitus. LID \u0026ndash;\u0026thinsp;10.3390/ijms25010342 [doi] LID \u0026ndash;\u0026thinsp;342.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZou, K. A. O. et al., \u003cem\u003eImpaired glucose partitioning in primary myotubes from severely obese women with type 2 diabetes.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhu, J. et al. Folate, Vitamin B6, and Vitamin B12 Status in Association With Metabolic Syndrome Incidence.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCummins, M. A. O., Delmonte, G., Wechsler, S. \u0026amp; Schlesinger, J. J. Alleviating mitochondrial dysfunction in diabetic cardiomyopathy through the Adipsin and Irak2 pathways.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou, Y. et al., \u003cem\u003eMetrnl Alleviates Lipid Accumulation by Modulating Mitochondrial Homeostasis in Diabetic Nephropathy.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCai, Y., Qi, X., Zheng, Y., Zhang, J. \u0026amp; Su, H. Lipid profile alterations and biomarker identification in type 1 diabetes mellitus patients under glycemic control.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFu, Y. et al., \u003cem\u003eCordycepin Ameliorates High Fat Diet-Induced Obesity by Modulating Endogenous Metabolism and Gut Microbiota Dysbiosis. LID \u0026ndash;\u0026thinsp;10.3390/nu16172859 [doi] LID \u0026ndash;\u0026thinsp;2859.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMonnerie, S. et al., \u003cem\u003eMetabolomic and Lipidomic Signatures of Metabolic Syndrome and its Physiological Components in Adults: A Systematic Review.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, L. et al., \u003cem\u003eMetabolic Homeostasis of Amino Acids and Diabetic Kidney Disease. LID \u0026ndash;\u0026thinsp;10.3390/nu15010184 [doi] LID \u0026ndash;\u0026thinsp;184.\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Diabetes Mellitus, Inflammation, Metabolite, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-6606669/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6606669/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eType 2 Diabetes Mellitus (T2DM) and obesity have reached epidemic levels globally. Although T2DM strain a heavy burden on population health, limited information is known about its metabolic signature, particularly in Emirati population.\u003c/p\u003e\u003cp\u003eWe explored Emiratis with controlled and uncontrolled T2DM metabolic profiles employing advanced machine-learning methods e.g. predictor randomization and bootstrap aggregation.\u003c/p\u003e\u003cp\u003eAlpha-ketoisovaleric acid, 2-pyrrolidinone, and uridine were identified as significant metabolic markers related to inflammation and metabolic stress between the two groups. Notable interactions between metabolites in the tryptophan metabolic pathway, including L-tryptophan and indole compounds, were observed, highlighting their potential role in T2DM progression. Alterations in Tricarboxylic Acid cycle intermediates and increased activation of Pentose Phosphate Pathway suggested adaptive responses to oxidative stress. Furthermore, metabolic changes were observed across the prediabetic, controlled, and uncontrolled diabetes stages, with metabolites such as 5-Dodecenoic acid and phosphatidylcholine identified as potential markers for distinguishing between these stages.\u003c/p\u003e\u003cp\u003eOur novel findings reveal the complex metabolic alterations associated with T2DM, providing an in-depth insight into the molecular level. These insights suggest that specific metabolites could serve as reliable biomarkers for early diagnosis and personalized treatment strategies. This approach could revolutionize the chronic conditions management, with effective and tailored interventions.\u003c/p\u003e","manuscriptTitle":"Metabolic Fingerprints of Diabetes: Machine Learning Reveals Distinct Biomarkers in Type 2 Diabetes Mellitus","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-04 11:04:59","doi":"10.21203/rs.3.rs-6606669/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e7958032-8a0e-4347-b8d7-c53fef854438","owner":[],"postedDate":"September 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":53816151,"name":"Health sciences/Endocrinology"},{"id":53816152,"name":"Health sciences/Medical research"},{"id":53816153,"name":"Health sciences/Molecular medicine"}],"tags":[],"updatedAt":"2026-02-23T11:40:03+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-04 11:04:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6606669","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6606669","identity":"rs-6606669","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}