An unsupervised learning approach to identify novel signatures of health and disease from multimodal data.
ABSTRACT: BACKGROUND:Modern medicine is rapidly moving towards a data-driven paradigm based on comprehensive multimodal health assessments. Integrated analysis of data from different modalities has the potential of uncovering novel biomarkers and disease signatures. METHODS:We collected 1385 data features from diverse modalities, including metabolome, microbiome, genetics, and advanced imaging, from 1253 individuals and from a longitudinal validation cohort of 1083 individuals. We utilized a combination of unsupervised machine learning methods to identify multimodal biomarker signatures of health and disease risk. RESULTS:Our method identified a set of cardiometabolic biomarkers that goes beyond standard clinical biomarkers. Stratification of individuals based on the signatures of these biomarkers identified distinct subsets of individuals with similar health statuses. Subset membership was a better predictor for diabetes than established clinical biomarkers such as glucose, insulin resistance, and body mass index. The novel biomarkers in the diabetes signature included 1-stearoyl-2-dihomo-linolenoyl-GPC and 1-(1-enyl-palmitoyl)-2-oleoyl-GPC. Another metabolite, cinnamoylglycine, was identified as a potential biomarker for both gut microbiome health and lean mass percentage. We identified potential early signatures for hypertension and a poor metabolic health outcome. Additionally, we found novel associations between a uremic toxin, p-cresol sulfate, and the abundance of the microbiome genera Intestinimonas and an unclassified genus in the Erysipelotrichaceae family. CONCLUSIONS:Our methodology and results demonstrate the potential of multimodal data integration, from the identification of novel biomarker signatures to a data-driven stratification of individuals into disease subtypes and stages-an essential step towards personalized, preventative health risk assessment.
Project description:Machine learning methods show promise to translate univariate biomarker findings into clinically useful multivariate decision support systems. At current, works in major depressive disorder have predominantly focused on neuroimaging and clinical predictor modalities, with genetic, blood-biomarker, and cardiovascular modalities lacking. In addition, the prediction of rehospitalization after an initial inpatient major depressive episode is yet to be explored, despite its clinical importance. To address this gap in the literature, we have used baseline clinical, structural imaging, blood-biomarker, genetic (polygenic risk scores), bioelectrical impedance and electrocardiography predictors to predict rehospitalization within 2 years of an initial inpatient episode of major depression. Three hundred and eighty patients from the ongoing 12-year Bidirect study were included in the analysis (rehospitalized: yes?=?102, no?=?278). Inclusion criteria was age ?35 and <66 years, a current or recent hospitalisation for a major depressive episode and complete structural imaging and genetic data. Optimal performance was achieved with a multimodal panel containing structural imaging, blood-biomarker, clinical, medication type, and sleep quality predictors, attaining a test AUC of 67.74 (p?=?9.99-05). This multimodal solution outperformed models based on clinical variables alone, combined biomarkers, and individual data modality prognostication for rehospitalization prediction. This finding points to the potential of predictive models that combine multimodal clinical and biomarker data in the development of clinical decision support systems.
Project description:Tumor biomarkers play important roles in tumor growth, invasion, and metastasis. Imaging of specific biomarkers will help to understand different biological activities, thereby achieving precise medicine for each head and neck squamous cell carcinoma (HNSCC) patient. Here, we describe various molecular targets and molecular imaging modalities for HNSCC imaging. An extensive search was undertaken in the PubMed database with the keywords including "HNSCC," "molecular imaging," "biomarker," and "multimodal imaging." Imaging targets in HNSCC consist of the epidermal growth factor receptor, cluster of differentiation 44 variant 6 (CD44v6), and mesenchymal-epithelial transition factor and integrins. Targeted molecular imaging modalities in HNSCC include optical imaging, ultrasound, magnetic resonance imaging, positron emission tomography, and single-photon emission computed tomography. Making the most of each single imaging method, targeted multimodal imaging has a great potential in the accurate diagnosis and therapy of HNSCC. By visualizing tumor biomarkers at cellular and molecular levels in vivo, targeted molecular imaging can be used to identify specific genetic and metabolic aberrations, thereby accelerating personalized treatment development for HNSCC patients.
Project description:Cognitive impairment is a feature of many psychiatric diseases, including schizophrenia. Here we aim to identify multimodal biomarkers for quantifying and predicting cognitive performance in individuals with schizophrenia and healthy controls. A supervised learning strategy is used to guide three-way multimodal magnetic resonance imaging (MRI) fusion in two independent cohorts including both healthy individuals and individuals with schizophrenia using multiple cognitive domain scores. Results highlight the salience network (gray matter, GM), corpus callosum (fractional anisotropy, FA), central executive and default-mode networks (fractional amplitude of low-frequency fluctuation, fALFF) as modality-specific biomarkers of generalized cognition. FALFF features are found to be more sensitive to cognitive domain differences, while the salience network in GM and corpus callosum in FA are highly consistent and predictive of multiple cognitive domains. These modality-specific brain regions define-in three separate cohorts-promising co-varying multimodal signatures that can be used as predictors of multi-domain cognition.
Project description:BACKGROUND:Although psychotherapy is one of the most efficacious and effective treatments for depression, limited accessibility to trained providers markedly limits access to care. In an attempt to overcome this obstacle, several platforms seeking to provide these services using digital modalities (eg, video, text, and chat) have been developed. However, the use of these modalities individually poses barriers to intervention access and acceptability. Multimodal platforms, comprising those that allow users to select from a number of available modalities, may be able to provide a solution to these concerns. OBJECTIVE:We aimed to investigate the preliminary effectiveness of providing psychotherapy through a multimodal digital psychotherapy platform. In addition, we aimed to examine differential responses to intervention by gender, self-reported physical health status, and self-reported financial status, as well as how prior exposure to traditional face-to-face psychotherapy affected the effectiveness of a multimodal digital psychotherapy intervention. Finally, we aimed to examine the dose-response effect. METHODS:Data were collected from a total of 318 active users of BetterHelp, a multimodal digital psychotherapy platform. Data on physical health status, financial status, and prior exposure to psychotherapy were obtained using self-report measures. Effectiveness was determined by the extent of symptom severity change, which was measured using the Patient Health Questionnaire at Time 1 (time of enrollment) and Time 2 (3 months after enrollment). Intervention dosage was measured as the sum of individual therapist-user interactions across modalities. RESULTS:Depression symptom severity was significantly reduced after the use of the multimodal digital psychotherapy intervention (P<.001). Individuals without prior traditional psychotherapy experience revealed increased improvement after intervention (P=.006). We found no significant dose-response effect of therapy, nor significant differences in outcomes across gender, self-reported financial status, and self-reported physical health status. CONCLUSIONS:Users of BetterHelp experienced significantly reduced depression symptom severity after engaging with the platform. Study findings suggest that this intervention is equally effective across gender, self-reported financial status, and self-reported physical health status and particularly effective for individuals without a history of psychotherapy. Overall, study results suggest that multimodal digital psychotherapy is a potentially effective treatment for adult depression; nevertheless, experimental trials are needed. We discuss directions for future research.
Project description:OBJECTIVE:To identify brain atrophy from structural-MRI and cerebral blood flow(CBF) patterns from arterial spin labeling perfusion-MRI that are best predictors of the A?-burden, measured as composite 18F-AV45-PET uptake, in individuals with early mild cognitive impairment(MCI). Furthermore, to assess the relative importance of imaging modalities in classification of A?+/A?- early mild cognitive impairment. METHODS:Sixty-seven ADNI-GO/2 participants with early-MCI were included. Voxel-wise anatomical shape variation measures were computed by estimating the initial diffeomorphic mapping momenta from an unbiased control template. CBF measures normalized to average motor cortex CBF were mapped onto the template space. Using partial least squares regression, we identified the structural and CBF signatures of A? after accounting for normal cofounding effects of age, sex, and education. RESULTS:18F-AV45-positive early-MCIs could be identified with 83% classification accuracy, 87% positive predictive value, and 84% negative predictive value by multidisciplinary classifiers combining demographics data, ApoE ?4-genotype, and a multimodal MRI-based A? score. INTERPRETATION:Multimodal-MRI can be used to predict the amyloid status of early-MCI individuals. MRI is a very attractive candidate for the identification of inexpensive and non-invasive surrogate biomarkers of A? deposition. Our approach is expected to have value for the identification of individuals likely to be A?+ in circumstances where cost or logistical problems prevent A? detection using cerebrospinal fluid analysis or A?-PET. This can also be used in clinical settings and clinical trials, aiding subject recruitment and evaluation of treatment efficacy. Imputation of the A?-positivity status could also complement A?-PET by identifying individuals who would benefit the most from this assessment.
Project description:Background: Among the neurodegenerative diseases of aging, sporadic Alzheimer's disease (AD) is the most prevalent and perhaps the most feared. With virtually no success at finding pharmaceutical therapeutics for altering progressive AD after diagnosis, research attention is increasingly directed at discovering biological and other markers that detect AD risk in the long asymptomatic phase. Both early detection and precision preclinical intervention require systematic investigation of multiple modalities and combinations of AD-related biomarkers and risk factors. We extend recent unbiased metabolomics research that produced a set of metabolite biomarker panels tailored to the discrimination of cognitively normal (CN), cognitively impaired and AD patients. Specifically, we compare the prediction importance of these panels with five other sets of modifiable and non-modifiable AD risk factors (genetic, lifestyle, cognitive, functional health and bio-demographic) in three clinical groups. Method: The three groups were: CN (n = 35), mild cognitive impairment (MCI; n = 25), and AD (n = 22). In a series of three pairwise comparisons, we used machine learning technology random forest analysis (RFA) to test relative predictive importance of up to 19 risk biomarkers from the six AD risk domains. Results: The three RFA multimodal prediction analyses produced significant discriminating risk factors. First, discriminating AD from CN was the AD metabolite panel and two cognitive markers. Second, discriminating AD from MCI was the AD/MCI metabolite panel and two cognitive markers. Third, discriminating MCI from CN was the MCI metabolite panel and seven markers from four other risk modalities: genetic, lifestyle, cognition and functional health. Conclusions: Salivary metabolomics biomarker panels, supplemented by other risk markers, were robust predictors of: (1) clinical differences in impairment and dementia and even; (2) subtle differences between CN and MCI. For the latter, the metabolite panel was supplemented by biomarkers that were both modifiable (e.g., functional) and non-modifiable (e.g., genetic). Comparing, integrating and identifying important multi-modal predictors may lead to novel combinations of complex risk profiles potentially indicative of neuropathological changes in asymptomatic or preclinical AD.
Project description:Glypican-1 (GPC-1) and other glypicans are a family of heparan sulfate proteoglycans. These proteins are highly expressed on the cell membrane and in the extracellular matrix, functioning mainly as modulators of growth factor signaling. Some of them are abnormally expressed in cancer, possibly involved in tumorigenesis, and detectable in blood as potential clinical biomarkers. GPC-1 is another glypican member that has been found to be associated with some cancers, and has increasingly interested the cancer field. Here we provide a brief review about GPC-1 in its expression, signaling and potential as a cancer biomarker.
Project description:An impediment to progress in mood disorders research is the lack of analytically valid and qualified diagnostic and treatment biomarkers. Consistent with the National Institute of Mental Health (NIMH)'s Research Domain Criteria (RDoC) initiative, the lack of diagnostic biomarkers has precluded us from moving away from a purely subjective (symptom-based) toward a more objective diagnostic system. In addition, treatment response biomarkers in mood disorders would facilitate drug development and move beyond trial-and-error toward more personalized treatments. As such, biomarkers identified early in the pathophysiological process are proximal biomarkers (target engagement), while those occurring later in the disease process are distal (disease pathway components). One strategy to achieve this goal in biomarker development is to increase efforts at the initial phases of biomarker development (i.e. exploration and validation) at single sites with the capability of integrating multimodal approaches across a biological systems level. Subsequently, resultant putative biomarkers could then undergo characterization and surrogacy as these latter phases require multisite collaborative efforts. We have used multimodal approaches - genetics, proteomics/metabolomics, peripheral measures, multimodal neuroimaging, neuropsychopharmacological challenge paradigms and clinical predictors - to explore potential predictor and mediator/moderator biomarkers of the rapid-acting antidepressants ketamine and scopolamine. These exploratory biomarkers may then be used for a priori stratification in larger multisite controlled studies during the validation and characterization phases with the ultimate goal of surrogacy. In sum, the combination of target engagement and well-qualified disease-related measures are crucial to improve our pathophysiological understanding, personalize treatment selection, and expand our armamentarium of novel therapeutics.
Project description:Type 2 diabetes (T2D) is associated with reduced gut microbiome diversity, although the cause is unclear. Metabolites generated by gut microbes also appear to be causative factors in T2D. We therefore searched for serum metabolites predictive of gut microbiome diversity in 1018 females from TwinsUK with concurrent metabolomic profiling and microbiome composition. We generated a Microbial Metabolites Diversity (MMD) score of six circulating metabolites that explained over 18% of the variance in microbiome alpha diversity. Moreover, the MMD score was associated with a significantly lower odds of prevalent (OR[95%CI] = 0.22[0.07;0.70], P = .01) and incident T2D (HR[95%CI] = 0.31[0.11,0.90], P = .03). We replicated our results in 1522 individuals from the ARIC study (prevalent T2D: OR[95%CI] = 0.79[0.64,0.96], P = .02, incident T2D: HR[95%CI] = 0.87[0.79,0.95], P = .003). The MMD score mediated 28%[15%,94%] of the total effect of gut microbiome on T2D after adjusting for confounders. Metabolites predicting higher microbiome diversity included 3-phenylpropionate(hydrocinnamate), indolepropionate, cinnamoylglycine and 5-alpha-pregnan-3beta,20 alpha-diol monosulfate(2) of which indolepropionate and phenylpropionate have already been linked to lower incidence of T2D. Metabolites correlating with lower microbial diversity included glutarate and imidazole propionate, of which the latter has been implicated in insulin resistance. Our results suggest that the effect of gut microbiome diversity on T2D is largely mediated by microbial metabolites, which might be modifiable by diet.
Project description:More than 1,000 different species of microbes have been found to live within the human oral cavity, where they play important roles in maintaining both oral and systemic health. Several studies have identified the core members of this microbial community; however, the factors that determine oral microbiome composition are not well understood. In this study, we exam the salivary oral microbiome of 1,049 Atlantic Canadians using 16S rRNA gene sequencing to determine which dietary, lifestyle, and anthropometric features play a role in shaping microbial community composition. Features that were identified as being significantly associated with overall composition then were additionally examined for genera, amplicon sequence variants, and predicted pathway abundances that were associated with these features. Several associations were replicated in an additional secondary validation data set. Overall, we found that several anthropometric measurements, including waist-hip ratio (WHR), height, and fat-free mass, as well as age and sex, were associated with overall oral microbiome structure in both our exploratory and validation data sets. We were unable to validate any dietary impacts on overall taxonomic oral microbiome composition but did find evidence to suggest potential contributions from factors such as the number of vegetable and refined grain servings an individual consumes. Interestingly, each one of these factors on its own was associated with only minor shifts in the overall taxonomic composition of the oral microbiome, suggesting that future biomarker identification for several diseases associated with the oral microbiome can be undertaken without the worry of confounding factors obscuring biological signals.IMPORTANCE The human oral cavity is inhabited by a diverse community of microbes, known as the human oral microbiome. These microbes play a role in maintaining both oral and systemic health and, as such, have been proposed to be useful biomarkers of disease. However, to identify these biomarkers, we first need to determine the composition and variation of the healthy oral microbiome. In this report, we investigate the oral microbiome of 1,049 healthy individuals to determine which genera and amplicon sequence variants are commonly found between individual oral microbiomes. We then further investigate how lifestyle, anthropometric, and dietary choices impact overall microbiome composition. Interestingly, the results from this investigation showed that while many features were significantly associated with oral microbiome composition, no single biological factor explained a variation larger than 2%. These results indicate that future work on biomarker detection may be encouraged by the lack of strong confounding factors.