PURPOSE: The aim was to determine the metabolic fingerprint that predicts warfarin response based on the international normalized ratio (INR) in patients who are already receiving warfarin (phase I: identification) and to ascertain the metabolic fingerprint that discriminates stable from unstable INR in patients starting treatment with warfarin (phase II: validation).
EXPERIMENTAL APPROACH: A total of 94 blood samples were collected for phase I: 44 patients with stable INR and 50 with unstable INR. Meanwhile, 23 samples were collected for phase II: nine patients with stable INR and 14 with unstable INR. Data analysis was performed using multivariate analysis including principal component analysis and partial least square-discriminate analysis (PLS-DA), followed by univariate and multivariate logistic regression (MVLR) to develop a model to identify unstable INR biomarkers.
KEY RESULTS: For phase I, the PLS-DA model showed the following results: sensitivity 93.18%, specificity 91.49% and accuracy 92.31%. In the MVLR analysis of phase I, ten regions were associated with unstable INR. For phase II, the PLS-DA model showed the following results: sensitivity 66.67%, specificity 61.54% and accuracy 63.64%.
CONCLUSIONS AND IMPLICATIONS: We have shown that the pharmacometabonomics technique was able to differentiate between unstable and stable INR with good accuracy. NMR-based pharmacometabonomics has the potential to identify novel biomarkers in plasma, which can be useful in individualizing treatment and controlling warfarin side effects, thus, minimizing undesirable effects in the future.
METHODS: Pre-dosed urine samples were collected from male Sprague-Dawley rats. The rats were treated with either LDA (10 mg/kg) or 1% methylcellulose (10 mL/kg) per oral for 28 days. The rats' stomachs were examined for gastric toxicity using a stereomicroscope. The urine samples were analyzed using a proton nuclear magnetic resonance spectroscopy. Metabolites were systematically identified by exploring established databases and multivariate analyses to determine the spectral pattern of metabolites related to LDA-induced gastric toxicity.
RESULTS: Treatment with LDA resulted in gastric toxicity in 20/32 rats (62.5%). The orthogonal projections to latent structures discriminant analysis (OPLS-DA) model displayed a goodness-of-fit (R2Y) value of 0.947, suggesting near-perfect reproducibility and a goodness-of-prediction (Q2Y) of -0.185 with perfect sensitivity, specificity and accuracy (100%). Furthermore, the area under the receiver operating characteristic (AUROC) displayed was 1. The final OPLS-DA model had an R2Y value of 0.726 and Q2Y of 0.142 with sensitivity (100%), specificity (95.0%) and accuracy (96.9%). Citrate, hippurate, methylamine, trimethylamine N-oxide and alpha-keto-glutarate were identified as the possible metabolites implicated in the LDA-induced gastric toxicity.
CONCLUSION: The study identified metabolic signatures that correlated with the development of a low-dose Aspirin-induced gastric toxicity in rats. This pharmacometabolomic approach could further be validated to predict LDA-induced gastric toxicity in patients with coronary artery disease.
METHODOLOGY: Eight (8) urine and serum samples each obtained from consenting healthy controls (HC), twenty-five (25) urine and serum samples each from first episode treatment naïve MDD (TNMDD) patients, and twenty (22) urine and serum samples each s from treatment naïve MDD patients 2 weeks after SSRI treatment (TWMDD) were analysed for metabolites using proton nuclear magnetic resonance (1HNMR) spectroscopy. The evaluation of patients' samples was carried out using Partial Least Squares Discriminant Analysis (PLS-DA) and Orthogonal Partial Least Square- Discriminant Analysis (OPLSDA) models.
RESULTS: In the serum, decreased levels of lactate, glucose, glutamine, creatinine, acetate, valine, alanine, and fatty acid and an increased level of acetone and choline in TNMDD or TWMDD irrespective of whether an OPLSDA or PLSDA evaluation was used were identified. A test for statistical validations of these models was successful.
CONCLUSION: Only some changes in serum metabolite levels between HC and TNMDD identified in this study have potential values in the diagnosis of MDD. These changes included decreased levels of lactate, glutamine, creatinine, valine, alanine, and fatty acid, as well as an increased level of acetone and choline in TNMDD. The diagnostic value of these changes in metabolites was maintained in samples from TWMDD patients, thus reaffirming the diagnostic nature of these metabolites for MDD.
MATERIALS AND METHODS: A literature search was carried out to gather eligible studies from the following widely sourced electronic databases such as Scopus, PubMed and Google Scholar using the combination of the following keywords: AD, MRS, brain metabolites, deep learning (DL), machine learning (ML) and artificial intelligence (AI); having the aim of taking the readers through the advancements in the usage of MRS analysis and related AI applications for the detection of AD.
RESULTS: We elaborate on the MRS data acquisition, processing, analysis, and interpretation techniques. Recommendation is made for MRS parameters that can obtain the best quality spectrum for fingerprinting the brain metabolomics composition in AD. Furthermore, we summarise ML and DL techniques that have been utilised to estimate the uncertainty in the machine-predicted metabolite content, as well as streamline the process of displaying results of metabolites derangement that occurs as part of ageing.
CONCLUSION: MRS has a role as a non-invasive tool for the detection of brain metabolite biomarkers that indicate brain metabolic health, which can be integral in the management of AD.