Figure - available from: Nature Medicine
This content is subject to copyright. Terms and conditions apply.
Study overview a, To learn metabolomic states from circulating blood metabolites, the eligible UK Biobank population (with NMR blood metabolomics and valid consent) was split into training, validation and test sets with 22-fold nested cross-validation based on the assigned UK Biobank assessment center. b, For each of the 22 partitions, the metabolomic state model was trained on the 168 metabolomic markers to predict metabolomic risk against 24 common disease endpoints. Subsequently, for each endpoint, CPH models were developed on the metabolomic state in combination with sets of commonly available clinical predictors to model disease risk. Predictions of the CPH model on the test set were aggregated for downstream analysis. c, The metabolomic state model was externally validated in four independent cohorts—the Whitehall II cohort and three from the BBMRI-NL consortium: the Rotterdam Study, the Leiden Longevity Study and the PROSPER cohort. d, In this study we consider clinical predictors from scores commonly applied in primary prevention. We additionally integrate variables into a comprehensive predictor set (PANEL) to investigate overlapping information with the metabolomic state. FH, family history.
Source publication
Risk stratification is critical for the early identification of high-risk individuals and disease prevention. Here we explored the potential of nuclear magnetic resonance (NMR) spectroscopy-derived metabolomic profiles to inform on multidisease risk beyond conventional clinical predictors for the onset of 24 common conditions, including metabolic,...
Citations
... For example, epigenetic aging biomarkers that are based on DNA methylation values at specific CpG sites have only weak correlations with metabolomics-based aging biomarkers, correlations ranging from −0.22 to 0.21 for MetaboAge and from 0 to 0.32 for MetaboHealth (Kuiper et al. 2023). The advantage of metabolomics-based aging biomarkers compared to other omics-based biomarkers lies in the fact that the metabolome carries more systemic information from multiple tissues across the body than, e.g., the methylome or transcriptome (Buergel et al. 2022). In addition, metabolomics-based biomarkers are trained on large sample sizes (Rutledge et al. 2022), and in recent years, nuclear magnetic resonance (NMR)-based metabolomics have matured and are now available at lower cost (Wishart et al. 2022). ...
Physical activity (PA) may delay the onset of age‐related diseases by decelerating biological aging. We investigated the association between leisure‐time physical activity (LTPA) and metabolomics‐based aging markers (MetaboAge and MetaboHealth) in late midlife and during 16 years of follow‐up. At the 16‐year follow‐up, we also investigated the association between device‐based PA and MetaboAge and MetaboHealth. We included 1816 individuals (mean age 61.6 years) from the Helsinki Birth Cohort Study at baseline and followed them up for 5 (n = 982) and 16 years (n = 744), respectively. LTPA was assessed via questionnaire at baseline and 16 years later and device‐based PA with ActiGraph accelerometer at the 16‐year follow‐up. Fasting blood samples were applied to calculate MetaboAge acceleration (ΔmetaboAge) and MetaboHealth at baseline and at both follow‐ups. Covariate‐adjusted multiple regression analyses and linear mixed models were applied to study the associations. A higher volume of LTPA at baseline was associated with a lower MetaboHealth score at the 5‐year follow‐up (p < 0.0001 for time × LTPA interaction). No associations were detected at the 16‐year follow‐up. An increase in LTPA over 16 years was associated with a decrease in MetaboHealth score (p < 0.001) and a decrease in LTPA with an increase in MetaboHealth score. Higher device‐based PA was associated with a lower MetaboHealth score, but not with ΔmetaboAge. In conclusion, higher LTPA in late midlife and device‐based PA in old age were associated with improved MetaboHealth. Increasing LTPA with age may protect against MetaboHealth‐based aging. The results support the importance of PA for biological aging in later life.
... A quintessential example of this is the SHapley Additive exPlanations (SHAP) (22,23), a model interpretation method, initially proposed by AI researchers and widely adopted across various scientific disciplines. For instance, SHAP has been used to quantify the contributions of individual metabolites to disease risk, identifying key metabolites influencing the risk of 24 investigated diseases (24). It has also been applied to interpret nanomaterial-plant-environment interactions, providing insights into the factors affecting the root uptake of metal-oxide nanoparticles and their interactions within the soil (25). ...
The rapid convergence of artificial intelligence (AI) with scientific research, often referred to as AI for Science (AI4Science), is reshaping the landscape of discovery across disciplines. Clarifying current progress and identifying promising pathways forward is essential to guide future development and unlock AI's transformative potential in scientific research. By analyzing AI-related research in leading natural and health science journals, we assess AI's integration into scientific fields and highlight opportunities for further growth. While AI's role in high-impact research is expanding, broader adoption remains hindered by cognitive and methodological gaps, necessitating targeted interventions to address these challenges. To accelerate AI4Science, we propose three key directions: equipping experimental scientists with user-friendly tools, developing proactive AI researchers within scientific workflows, and fostering a thriving AI-Science ecosystem. Additionally, we introduce a structured AI4Science workflow to guide both experimental scientists and AI researchers in leveraging AI for discovery, while proposing strategies to overcome adoption barriers. Ultimately, this work aims to drive broader AI integration in research, advancing scientific discovery and innovation across disciplines.
... BCAA are involved in synthesis of neurotransmitters, proteins and energy production, therefore their effects reach farther than the described here and have greater implication in AD (47). Lower blood levels of BCAA were a main contributor to predicted risk of dementia and AD in several cohorts, even 10 years before disease onset (48)(49)(50)(51). On the contrary, high brain BCAA levels were associated with AD pathology and cognitive impairment (27), suggesting the importance of their utilization by brain. ...
Alzheimer's disease (AD) risk and progression are significantly influenced by APOE genotype with APOE4 increasing and APOE2 decreasing susceptibility compared to APOE3. While the effect of those genotypes was extensively studied on blood metabolome, less is known about their impact in the brain. Here we investigated the impacts of APOE genotypes and aging on brain metabolic profiles across the lifespan, using human APOE-targeted replacement mice. Biocrates P180 targeted metabolomics platform was used to measure a broad range of metabolites probing various metabolic processes. In all genotypes investigated we report changes in acylcarnitines, biogenic amines, amino acids, phospholipids and sphingomyelins during aging. The decreased ratio of medium to long-chain acylcarnitine suggests a reduced level of fatty acid β-oxidation and thus the possibility of mitochondrial dysfunction as these animals age. Additionally, aging APOE2/2 mice had altered branch-chain amino acids (BCAA) profile and increased their downstream metabolite C5 acylcarnitine, indicating increased branched-chain amino acid utilization in TCA cycle and better energetic profile endowed by this protective genotype. We compared these results with human dorsolateral prefrontal cortex metabolomic data from the Religious Orders Study/Memory and Aging Project, and we found that the carriers of APOE2/3 genotype had lower markers of impaired BCAA katabolism, including tiglyl carnitine, methylmalonate and 3-methylglutaconate. In summary, these results suggest a potential involvement of the APOE2 genotype in BCAA utilization in the TCA cycle and nominate these humanized APOE mouse models for further study of APOE in AD, brain aging, and brain BCAA utilization for energy. We have previously shown lower plasma BCAA to be associated with incident dementia, and their higher levels in brain with AD pathology and cognitive impairment. Those findings together with our current results could potentially explain the AD-protective effect of APOE2 genotype by enabling higher utilization of BCAA for energy during the decline of fatty acid β-oxidation.
... 27 Thus, we imputed missing data for included variables using a machine learning algorithm called random forest imputation implemented in the R package "missForest", 28 which was widely used in medical research. 29,30 All included variables contained less than 30% missing values. ...
Background
Epidemiological studies suggest that elevated blood urea nitrogen (BUN) and reduced serum albumin could independently predict adverse clinical outcomes in patients with chronic obstructive pulmonary disease (COPD). However, the predictive performance of BUN-albumin ratio (BAR) in critically ill patients with COPD remains to be confirmed. This study aimed to investigate the association between BAR and all-cause mortality in intensive care unit (ICU) patients with COPD.
Methods
This was a retrospective study that included COPD patients with BUN and serum albumin value on the first day of each ICU admission and data were obtained from the eICU Collaborative Research Database. The included COPD patients were divided into three groups stratified by BAR tertiles (T1-T3). Multivariate logistic regression and Cox proportional hazards models were used to examine the association between BAR and all-cause in-hospital and ICU mortality, respectively. Kaplan–Meier curves were plotted to evaluate survival differences among three groups and discrepancies were compared with the log–rank test.
Results
A total of 4037 patients were included in the final analysis and the in-hospital and ICU mortality rates were 11.79% and 6.51%, respectively. The multivariate logistic regression analyses showed that continuous BAR was a significant risk predictor of in-hospital mortality (OR: 1.039, 95% CI: 1.026–1.052, P < 0.001) and ICU mortality (OR: 1.030, 95% CI: 1.015–1.045, P < 0.001) in fully adjusted model. The Cox proportional hazards models revealed that patients in the highest BAR tertile (T3) were significantly associated with higher risk of in-hospital mortality (HR: 1.983, 95% CI: 1.419–2.772, P < 0.001) and ICU mortality (HR: 2.166, 95% CI: 1.373–3.418, P < 0.001). The Kaplan–Meier curves showed that the survival differences of all-cause mortality were statistically significant in three tertile groups (log-rank P < 0.0001). Correlated subgroup analyses indicated that this positive association might vary in certain population settings.
Conclusion
High level of BAR is associated with the increasing all-cause mortality in critically ill patients with COPD. As an innovative and promising biomarker, BAR might be useful in predicting high risk of death in patients with COPD.
... Omics sciences represent a very promising instrument to perform the analysis of patients and their biological characteristics within the dynamic context of disease evolution, thus enabling the molecular characterization of a disease onset and evolution, and providing insight into individual susceptibility to drug treatments [5][6][7][8][9][10]. Given these premises, metabolomics and lipoproteomics present themselves as compelling approaches for investigating alterations of multiple biochemical networks throughout the entire course of AD [11][12][13][14][15][16][17][18]. ...
Background
Alzheimer’s disease (AD) is the most frequent neurodegenerative disorder worldwide. The great variability in disease evolution and the incomplete understanding of the molecular mechanisms underlying AD make it difficult to predict when a patient will convert from prodromal stage to dementia. We hypothesize that metabolic alterations present at the level of the brain could be reflected at a systemic level in blood serum of patients, and that these alterations could be used as prognostic biomarkers.
Methods
This pilot study proposes a serum investigation via nuclear magnetic resonance (NMR) spectroscopy in a consecutive series of AD patients including 57 patients affected by Alzheimer’s disease at dementia stage (AD-dem) and 45 patients with mild cognitive impairment (MCI) due to AD (MCI-AD). As control group, we considered 31 subjects with mild cognitive impairment in whom AD and other neurodegenerative disorders were excluded (MCI). A panel of 26 metabolites and 112 lipoprotein-related parameters was quantified and the logistic LASSO regression algorithm was employed to identify the optimal combination of metabolites-lipoproteins and their ratios to discriminate the groups of interest.
Results
In the training set, our model classified AD-dem and MCI with an accuracy of 81.7%. These results were reproduced in the validation set (accuracy 75.0%). Evolution of MCI-AD patients was evaluated over time. Patients who displayed a decrease in MMSE < 1.5 point per year were considered at lower progression rate: we obtained a division in 18 MCI-AD at lower progression rate (MCI-AD LR) and 27 at higher progression rate (MCI-AD HR). The model calculated using 4 metabolic features identified MCI-AD LR and MCI-AD HR with an accuracy of 73.3%.
Conclusions
The identification of potential novel peripheral biomarkers of Alzheimer’s disease, as proposed in this study, opens a new prospect for an innovative and minimally invasive method to identify AD in its very early stages. We proposed a novel approach able to sub-stratify MCI-AD patients identifying those associated with a faster rate of clinical progression.
... A notable example is a study using the UK Biobank data, which developed a deep residual multitask neural network for blood metabolome analysis. This network effectively discerns the metabolomic profiles of 24 common diseases [14]. Recent research has increasingly applied metabolomics to OP and ON, identifying distinct metabolic variations in affected individuals [15,16]. ...
Background
Osteopenia (ON) and osteoporosis (OP) are highly prevalent among postmenopausal women and poses a challenge for early diagnosis. Therefore, identifying reliable biomarkers for early prediction using metabolomics is critically important.
Methods
Initially, non-targeted metabolomics was employed to identify differential metabolites in plasma samples from cohort 1, which included healthy controls (HC, n = 23), osteonecrosis (ON, n = 36), and osteoporosis (OP, n = 37). Subsequently, we performed targeted metabolomic validation of 37 amino acids and their derivatives in plasma samples from cohort 2, consisting of healthy controls (HC, n = 10), osteonecrosis (ON, n = 10), and osteoporosis (OP, n = 10).
Results
The non-targeted metabolomic analysis revealed an increase in differential metabolites with the progression of the disease, showing abnormalities in lipid and organic acid metabolism in ON and OP patients. Several substances were found to correlate positively or negatively with bone mineral density (BMD), for example, N-undecanoylglycine, sphingomyelins, and phosphatidylinositols exhibited positive correlations with BMD, while acetic acid, phenylalanine, taurine, inosine, and pyruvic acid showed negative correlations with BMD. Subsequently, targeted validation of 37 amino acids and their metabolites revealed six amino acids related to ON and OP.
Conclusion
Significant metabolomic features were identified between HC and patients with ON/OP, with multiple metabolites correlating positively or negatively with BMD. Integrating both targeted and non-targeted metabolomic results suggests that lipid, organic acid, and amino acid metabolism may represent important metabolomic characteristics of patients with OP, offering new insights into the development of metabolomic applications in OP.
... Beyond the well-established metabolic basis of multisystem diseases [58][59][60] , recent studies have confirmed connections between circulatory metabolome and retinal health 22,24,29,30,46,61 . Inspired by prior work bridging the mutual link of the retina with both metabolic factors and systemic health 12,15,28,12,15,28 , the photoreceptor layer's broad implications across multiple systems might stem from specific circulating metabolic stresses. ...
... A high-throughput nuclear magnetic resonance (NMR) platform (Nightingale Health, Finland) was used to quantify the metabolite concentrations from the plasma samples collected from participants in both cohorts 59,82,83 . Sample collection was undertaken at baseline in local assessment centers across the UK between 2007 and 2010 from UKB participants, and in Zhongshan Ophthalmic Center between 2017 and 2021 from GDES participants. ...
... CPH models assessed the associations between PMW and various outcome risks, and interaction analyses assessed the effects of age, sex, ethnicity, deprivation, and educational attainment on the associations, with the same covariates adjusted as above. Participants were stratified based on predicted outcome-specific PMW states, and the cumulative event rates were compared across the top, middle, and bottom 10% of states for each specific outcome 12,59 . C-statistics were calculated to assess PMW's predictivity for multisystem outcomes, and its predictive value was then compared with that of 10 conventional predictors (age, sex, deprivation, income, smoking, drinking, educational attainment, body mass index, and use of lipid-lowering and antihypertensive medication) in both cohorts. ...
Photoreceptors are specialized neurons at the core of the retina’s functionality, with optical accessibility and exceptional sensitivity to systemic metabolic stresses. Here we show the ability of risk-free, in vivo photoreceptor assessment as a window into systemic health and identify shared metabolic underpinnings of photoreceptor degeneration and multisystem health outcomes. A thinner photoreceptor layer thickness is significantly associated with an increased risk of future mortality and 13 multisystem diseases, while systematic analyses of circulating metabolomics enable the identification of 109 photoreceptor-related metabolites, which in turn elevate or reduce the risk of these health outcomes. To translate these insights into a practical tool, we developed an artificial intelligence (AI)-driven photoreceptor metabolic window framework and an accompanying interpreter that comprehensively captures the metabolic landscape of photoreceptor–systemic health linkages and simultaneously predicts 16 multisystem health outcomes beyond established approaches while retaining interpretability. Our work, replicated across cohorts of diverse ethnicities, reveals the potential of photoreceptors to inform systemic health and advance a multisystem perspective on human health by revealing eye–body connections and shared metabolic influences.
... 2,3 Markers of metabolic activity, such as lipoproteins, cholesterol, fatty acids, and glucose, are now wellcaptured in plasma by high-throughput nuclear magnetic resonance (NMR) spectroscopy, 4 explaining more variance in metabolic dysfunction than traditional measurements. 5 As the world's population is becoming progressively older and the prevalence of obesity is rising, 6 a better understanding of the connection between metabolic markers and the brain will help to develop more specific interventions to combat the these global public health challenges. ...
Background: Metabolic processes form the basis of the development, functioning and maintenance of the brain. Despite accumulating evidence of the vital role of metabolism in brain health, no study to date has comprehensively investigated the link between circulating markers of metabolic activity and in vivo brain morphology in the general population. Methods: We performed uni- and multivariate regression on metabolomics and MRI data from 24,940 UK Biobank participants, to estimate the individual and combined associations of 249 circulating metabolic markers with 91 measures of global and regional cortical thickness, surface area and subcortical volume. We investigated similarity of the identified spatial patterns with brain maps of neurotransmitters, and used Mendelian randomization to uncover causal relationships between metabolites and the brain. Results: Intracranial volume and total surface area were highly significantly associated with circulating lipoproteins and glycoprotein acetyls, with correlations up to .15. There were strong regional associations of individual markers with mixed effect directions, with distinct patterns involving frontal and temporal cortical thickness, brainstem and ventricular volume. Mendelian randomization provided evidence of bidirectional causal effects, with the majority of markers affecting frontal and temporal regions. Discussion: The results indicate strong bidirectional causal relationships between circulating metabolic markers and distinct patterns of global and regional brain morphology. The generated atlas of associations provides a better understanding of the role of metabolic pathways in structural brain development and maintenance, in both health and disease.
... Importantly, our approach, based on routine health records, shows large discriminative improvements for the majority of diseases compared with conventionally tested biomarkers [55][56][57] and can be generalized across diverse health systems, populations, and ethnicities. However, we also see that including the medical history over age and sex deteriorated the performance for a subset of 0.7% (UK Biobank) and 5.5% (All Of Us cohort), respectively. ...
The COVID-19 pandemic exposed a global deficiency of systematic, data-driven guidance to identify high-risk individuals. Here, we illustrate the utility of routinely recorded medical history to predict the risk for 1741 diseases across clinical specialties and support the rapid response to emerging health threats such as COVID-19. We developed a neural network to learn from health records of 502,489 UK Biobank participants. Importantly, we observed discriminative improvements over basic demographic predictors for 1546 (88.8%) endpoints. After transferring the unmodified risk models to the All of US cohort, we replicated these improvements for 1115 (78.9%) of 1414 investigated endpoints, demonstrating generalizability across healthcare systems and historically underrepresented groups. Ultimately, we showed how this approach could have been used to identify individuals vulnerable to severe COVID-19. Our study demonstrates the potential of medical history to support guidance for emerging pandemics by systematically estimating risk for thousands of diseases at once at minimal cost.
... 167 Beyond genomic approaches, other -omics methodologies have demonstrated the ability to stratify carotid atherosclerotic plaque profiles. 162,[168][169][170][171][172] The large data sets produced by omics approaches can also be integrated with machine learning techniques for data analysis. Chen et al. employed a neural network algorithm and hierarchical clustering to identify different subclinical carotid atherosclerotic endotypes related to different risk profiles of future cardiovascular events in two independent populations. ...
Cardiovascular disease remains a prominent cause of disability and premature death worldwide. Within this spectrum, carotid artery atherosclerosis is a complex and multifaceted condition, and a prominent precursor of acute ischaemic stroke and other cardiovascular events. The intricate interplay among inflammation, oxidative stress, endothelial dysfunction, lipid metabolism, and immune responses participates in the development of lesions, leading to luminal stenosis and potential plaque instability. Even non-stenotic plaques can precipitate a sudden cerebrovascular event, regardless of the degree of luminal encroachment. In this context, carotid imaging modalities have proved their efficacy in providing in vivo characterization of plaque features, contributing substantially to patient risk stratification and clinical management. This review emphasizes the importance of identifying high-risk individuals by use of current imaging modalities, biomarkers, and risk stratification tools. Such approaches inform early intervention and the implementation of personalized therapeutic strategies, ultimately enhancing patient outcomes in the realm of cardiovascular disease management.
