Multivariate Receiver Operating Characteristic Analyses Research Articles

Serum proteomic profiling reveals potential inflammatory biomarkers in long-COVID patients: a comparative analysis with healthy controls.

The highly transmissible severe acute respiratory syndrome-Coronavirus-2 was responsible for the 2020 COVID-19 pandemic. COVID-19 mostly affects the respiratory system; however, this infection also affects several other organs. In addition, the sequelae of this disease affect patients for several months after recovery, resulting in long-COVID syndrome. In order to characterize the differences between healthy control individuals and long-COVID patients, proteomic profiling of the serum of both groups was performed by mass spectrometry. The obtained data were analyzed with multivariate and univariate statistical analyses. Initially, performing a partial latent square discriminant analysis (PLS-DA) made it possible to identify thirty-three proteins of interest, which were then subjected to a receiver operating characteristic (ROC) analysis. Four proteins were identified as potential stand-alone biomarkers: Sirtuin 1, Natriuretic Peptide B, Hemopexin, and Arachidonate 5-Lipoxygenase. Moreover, a multivariate ROC analysis identified a panel of biomarkers composed of Natriuretic Peptide B, Anterior Gradient 2 Protein, Adiponectin, Endothelin Converting Enzyme 1, Interferon Induced Transmembrane Protein 1, Mannose Binding Lectin 2, Prostaglandin-Endoperoxide Synthase 2, Pirin, Prostaglandin Reductase 1 and Cystatin C. The identified biomarkers are associated with inflammatory processes, corroborating literature evidence that long-COVID patients develop an inflammatory state that damages many tissues. Nevertheless, these data should be validated in a larger cohort.

Virtual diagnosis of diabetic nephropathy using metabolomics in place of kidney biopsy: The DIAMOND study

AimsTo explore the clinical factors and urinary metabolites that predict biopsy-confirmed diabetic nephropathy (DN) in patients with type 2 diabetes mellitus (T2DM). MethodsData from the medical records of 126 patients with T2DM who underwent kidney biopsy between January 2010 and October 2020 at a single-center were retrospectively reviewed to investigate the clinical factors that predict DN. Urine samples were collected to perform urine metabolomics in patients with T2DM divided by biopsy-confirmed DN, immunoglobulin A, and membranous nephropathy, and a control group of healthy participants. Each group comprised 11 age- and sex-matched participants. A prediction model was developed using a combination of clinical factors and urinary metabolites, and a multivariate receiver operating characteristic (ROC) analysis was conducted. ResultsAge, presence of proliferative diabetic retinopathy, T2DM duration, and hemoglobin A1c levels were clinical factors predictive of DN. Four urinary metabolites (alanine, choline, N-phenylacetylglycine, and trigonelline) had variable importance in projection scores > 1 and were predictive of DN. When conducting multivariate ROC analysis with a combination of clinical factors and urinary metabolites, the area under the curve was 1.000. ConclusionsThe combination of clinical factors and urinary metabolites is highly valuable for predicting biopsy-confirmed DN in patients with T2DM.

Serum and urine metabolomic biomarkers for predicting prognosis in patients with immunoglobulin A nephropathy.

Immunoglobulin A nephropathy (IgAN) is the most prevalent form of glomerulonephritis worldwide. Prediction of disease progression in IgAN can help to provide individualized treatment based on accurate risk stratification. We performed proton nuclear magnetic resonance-based metabolomics analyses of serum and urine samples from healthy controls, non-progressor (NP), and progressor (P) groups to identify metabolic profiles of IgAN disease progression. Metabolites that were significantly different between the NP and P groups were selected for pathway analysis. Subsequently, we analyzed multivariate area under the receiver operating characteristic (ROC) curves to evaluate the predictive power of metabolites associated with IgAN progression. We observed several distinct metabolic fingerprints of the P group involving the following metabolic pathways: glycolipid metabolism; valine, leucine, and isoleucine biosynthesis; aminoacyl-transfer RNA biosynthesis; glycine, serine, and threonine metabolism; and glyoxylate and dicarboxylate metabolism. In multivariate ROC analyses, the combinations of serum glycerol, threonine, and proteinuria (area under the curve [AUC], 0.923; 95% confidence interval [CI], 0.667-1.000) and of urinary leucine, valine, and proteinuria (AUC, 0.912; 95% CI, 0.667-1.000) showed the highest discriminatory ability to predict IgAN disease progression. This study identified serum and urine metabolites profiles that can aid in the identification of progressive IgAN and proposed perturbed metabolic pathways associated with the identified metabolites.

3 CpG Methylation Biomarkers for the Diagnosis of Polycystic Ovary Syndrome (PCOS) in Blood Samples.

Polycystic ovary syndrome (PCOS) is a common endocrine disease in women that seriously interferes with patient's metabolic and reproductive functions. The current diagnostic criteria for PCOS are expert-based and still disputed. Previous studies have identified changes in DNA methylation in peripheral blood of women with PCOS, but their diagnostic potential for PCOS remains to be studied. The present study aimed to identify potential methylation biomarkers for the diagnosis of PCOS in blood. Methylation profiling of peripheral blood was downloaded from a public database, Gene Expression Omnibus (GEO), including 30 PCOS patients (diagnosed with the revised 2003 Rotterdam consensus criteria) and 30 age-matched healthy women recruited from Centre of Reproductive Medicine, Linyi People's Hospital, Shandong, China. Weighted gene co-expression network analysis (WGCNA) was utilized to identify PCOS-related co-methylation CpG sites (co- MPs). Functional enrichment analysis was performed on the localized genes of PCOS-related co- MPs. The least absolute shrinkage and selection operator (LASSO) regression was used to screen out CpG methylation signatures for PCOS diagnosis, and receiver operating characteristic (ROC) analysis was conducted to evaluate their diagnostic accuracy. To assess the accuracy of the combination of the investigated indicators, multivariate ROC analysis was performed on the predicted probability values obtained using binary logistic regression on the methylation levels of selected CpGs. Seven co-methylation modules were obtained, among which the turquoise module is the most relevant to PCOS, containing 194 co-MPs. The genes that these co-MPs located in were mainly associated with the immune-related pathway. According to LASSO regression, three Co- MPs (cg23464743, cg06834912, cg00103771) were identified as potential diagnostic biomarkers of PCOS. ROC analysis showed an AUC (area under curve) of 0.7556 (sensitivity 60.0%, specificity 83.3%) for cg23464743, 0.7822 (sensitivity 70.0%, specificity 80.0%) for cg06834912, and 0.7611 (sensitivity 63.3%, specificity 83.3%) for cg00103771. The diagnostic accuracy of the combination of these 3 indicators presented to be higher than any single one of them, with the AUC of 0.8378 (sensitivity 73.3%, specificity 93.3%). The combination of 3 CpG methylation signatures in blood was identified with a good diagnostic accuracy for PCOS, which may bring new insight into the development of PCOS diagnostic markers in the future.

Ovarian Cancer Detection Using Targeted Plasma Metabolic Profiling

Ovarian cancer (OC) is the second most common form of gynecological cancer and is the most fatal among all forms of gynecological malignancies. Despite the important role of metabolic processes in the molecular pathogenesis of OC, robust metabolic markers to enable effective screening, rapid diagnosis, accurate surveillance, and therapeutic monitoring of OC are still lacking. In this study, we present a targeted liquid chromatography‐tandem mass spectrometry (LC‐MS/MS)‐based metabolic profiling approach for the identification of metabolic biomarker candidates that could enable expedited, highly sensitive and specific OC detection. Using this targeted approach, 90 plasma metabolites from many metabolic pathways of potential biological significance were reliably detected and monitored in 218 samples taken from the three groups of subjects (78 OC patients, 50 benign samples, and 90 healthy controls). Multivariate significance testing and receiver operating characteristic (ROC) analysis revealed 8 significant metabolites (p<0.05) with high magnitudes of fold change (FC>2) and high predictive accuracy [area under the curve (AUC)>0.90] for distinguishing healthy controls from OC patients. For discrimination of OC patients from benign samples, the results of our model refinement pipeline informed the construction of a logistic regression classifier defined by a 5‐metbaolite panel of potential plasma biomarkers, exhibiting roughly 75% predictive accuracy using a 50% random‐split training and testing paradigm for external validation. Multivariate ROC analysis showed enhanced classification performance for identification of OC patients from benign samples, relative to individual metabolites. Pathway analysis revealed significant disturbances in glycine, serine, and threonine metabolism; glyoxylate and dioxylate metabolism; the pentose phosphate pathway; and histidine metabolism. The results expand basic knowledge of the metabolome related to OC pathogenesis relative to both healthy and benign samples, revealing potential pathways or markers that can be targeted therapeutically. This study also provides a promising basis for the development of larger multi‐site projects to validate our findings across population groups and further advances the development of improved clinical care for OC patients.

Potential Salivary mRNA Biomarkers for Early Detection of Oral Cancer.

We evaluated potential biomarkers in human whole saliva for the early diagnosis of oral squamous cell carcinoma (OSCC). We selected 30 candidate genes with relevance to cancer from recent reports in PubMed. Saliva samples were obtained from 34 non-tumor control and 33 OSCC patients. Real-time PCR was performed, and mRNA levels were compared. Normalized mRNA levels of six genes (NGFI-A binding protein 2 (NAB2), cytochrome P450, family 27, subfamily A, polypeptide 1 (CYP27A1), nuclear pore complex interacting protein family, member B4 (NPIPB4), monoamine oxidase B (MAOB), sialic acid acetyltransferase (SIAE), and collagen, type III, alpha 1 (COL3A1)) were significantly lower in saliva of OSCC patients. Receiver operating characteristics (ROC) analysis was used to individually evaluate the predictive power of the potential biomarkers for OSCC diagnosis. The area under the curve (AUC) values were evaluated for the OSCC vs. non-tumor groups via univariate ROC analyses, as well as multivariate ROC analyses of combinations of multiple potential biomarkers. The combination of CYP27A1 + SIAE showed a favorable AUC value of 0.84. When we divided saliva samples into two groups according to age using a 60-year cut-off, with OSCC patients and controls evaluated together, the AUC of MAOB–NAB2 was more predictive of OSCC in the under-60 group (AUC, 0.91; sensitivity, 0.92; and specificity, 0.86) than any other gene combination. These results are expected to aid the early diagnosis of OSCC, especially in patients under 60 years of age. While more studies with larger numbers of patients are necessary, our result suggest that salivary mRNA would be a potent biomarker for early OSCC diagnosis.

Quantitation of in vivo brain glutathione conformers in cingulate cortex among age-matched control, MCI, and AD patients using MEGA-PRESS.

Oxidative stress (OS) plays an important role in Alzheimer's disease (AD) and glutathione (GSH) mitigates this effect by maintaining redox‐imbalance and free‐radical neutralization. Quantified brain GSH concentration provides distinct information about OS among age‐matched normal control (NC), mild cognitive impairment (MCI) and AD patients. We report alterations of in vivo GSH conformers, along with the choline, creatine, and N‐acetylaspartate levels in the cingulate cortex (CC) containing anterior (ACC) and posterior (PCC) regions of 64 (27 NC, 19 MCI, and 18 AD) participants using MEscher–GArwood‐Point‐RESolved spectroscopy sequence. Result indicated, tissue corrected GSH depletion in PCC among MCI (p = .001) and AD (p = .028) and in ACC among MCI (p = .194) and AD (p = .025) as compared to NC. Effects of the group, region, and group × region on GSH with age and gender as covariates were analyzed using a generalized linear model with Bonferroni correction for multiple comparisons. A significant effect of group with GSH depletion in AD and MCI was observed as compared to NC. Receiver operator characteristic (ROC) analysis of GSH level in CC differentiated between MCI and NC groups with an accuracy of 82.8% and 73.5% between AD and NC groups. Multivariate ROC analysis for the combined effect of the GSH alteration in both ACC and PCC regions provided improved diagnostic accuracy of 86.6% for NC to MCI conversion and 76.4% for NC to AD conversion. We conclude that only closed GSH conformer depletion in the ACC and PCC regions is critical and constitute a potential biomarker for AD.

A novel multiple biomarker panel for the early detection of high-grade serous ovarian carcinoma

IntroductionSince the majority of patients are diagnosed at an advanced stage, ovarian cancer remains the most lethal gynecologic malignancy. There is no single biomarker with the sensitivity and specificity required for effective cancer screening; therefore, we investigated a panel of novel biomarkers for the early detection of high-grade serous ovarian carcinoma. MethodsTwelve serum biomarkers with high differential gene expression and validated antibodies were selected: IL-1Ra, IL-6, Dkk-1, uPA, E-CAD, ErbB2, SLPI, HE4, CA125, LCN2, MSLN, and OPN. They were tested using Simple Plex™, a multi-analyte immunoassay platform, in samples collected from 172 patients who were either healthy, had benign gynecologic pathologies, or had high-grade serous ovarian adenocarcinomas. The receiver operating characteristic (ROC) curve, ROC area under the curve (AUC), and standard error (SE) of the AUC were obtained. Univariate ROC analyses and multivariate ROC analyses with the combination of multiple biomarkers were performed. ResultsThe 4-marker panel consisting of CA125, HE4, E-CAD, and IL-6 had the highest ROC AUC. When evaluated for the ability to distinguish early stage ovarian cancer from a non-cancer control, not only did this 4-marker panel (AUC=0.961) performed better than CA 125 alone (AUC=0.851; P=0.0150) and HE4 alone (AUC=0.870; P=0.0220), but also performed significantly better than the 2- marker combination of CA125+HE4 (AUC=0.922; P=0.0278). The 4-marker panel had the highest average sensitivity under the region of its ROC curve corresponding to specificity ranging from 100% down to ~95%. ConclusionThe four-marker panel, CA125, HE4, E-CAD, and IL-6, shows potential in detecting serous ovarian cancer at earlier stages. Additional validation studies using the biomarker combination in ovarian cancer patients are warranted.

Changes in quantitative CT image features of ground-glass nodules in differentiating invasive pulmonary adenocarcinoma from benign and in situ lesions: histopathological comparisons

To evaluate progressive changes in quantitative CT features of the non-solid component of ground-glass nodules (GGNs) from baseline to follow-up to differentiate invasive (minimally invasive adenocarcinoma [MIA] and invasive adenocarcinoma [IA]) GGNs from benign or pre-invasive (adenocarcinoma in situ [AIS]) lesions. Patients with a GGN detected at baseline and follow-up computed tomography (CT), examined by tissue sampling were included in the study. The diameter and mean, maximum, minimum CT density and density deviation from the non-solid component of whole GGNs were measured. Progression of these features over time was analysed by linear regression analysis. Multivariate receiver operating characteristics analyses of combined measures created by a logistic regression model were performed to evaluate diagnostic performance for invasive GGNs. Sixty-one patients (24 males) with 70 GGNs were included. Fifteen GGNs were benign, six AIS, 38 MIA, and 11 IA. The mean diameter of all histological subtypes increased from baseline to follow-up, the largest increase was found in the MIA group (p<0.001). For MIA and IA, the mean, maximum, minimum density, and density deviation had a positive correlation over time, whilst benign and pre-invasive GGNs showed a negative correlation for these features. A diagnostic model based on three GGN features (increasing in diameter, mean density, and density deviation) identified invasive GGNs with a sensitivity, specificity and area under the receiver operating characteristic (ROC) curve (AUC) of 83.7%, 61.9%, and 0.786, respectively (p<0.001). In GGN follow-up, the diameter of benign and AIS, and invasive GGNs significantly increased. Additional analysis of mean density and density deviation in the non-solid component may help to identify invasive GGNs.

Accuracy of MR markers for differentiating Progressive Supranuclear Palsy from Parkinson's disease.

BackgroundAdvanced brain MR techniques are useful tools for differentiating Progressive Supranuclear Palsy from Parkinson's disease, although time-consuming and unlikely to be used all together in routine clinical work. We aimed to compare the diagnostic accuracy of quantitative morphometric, volumetric and DTI metrics for differentiating Progressive Supranuclear Palsy-Richardson's Syndrome from Parkinson's disease.Methods23 Progressive Supranuclear Palsy-Richardson's Syndrome and 42 Parkinson's disease patients underwent a standardized 1.5T brain MR protocol comprising high-resolution T1W1 and DTI sequences. Brainstem and cerebellar peduncles morphometry, automated volumetric analysis of brain deep gray matter and DTI metric analyses of specific brain structures were carried out. We determined diagnostic accuracy, sensitivity and specificity of MR-markers with respect to the clinical diagnosis by using univariate receiver operating characteristics curve analyses. Age-adjusted multivariate receiver operating characteristics analyses were then conducted including only MR-markers with a sensitivity and specificity exceeding 80%.ResultsMorphometric markers (midbrain area, pons to midbrain area ratio and MR Parkinsonism Index), DTI parameters (infratentorial structures) and volumetric analysis (thalamus, putamen and pallidus nuclei) presented moderate to high diagnostic accuracy in discriminating Progressive Supranuclear Palsy-Richardson's Syndrome from Parkinson's disease, with midbrain area showing the highest diagnostic accuracy (99%) (mean ± standard deviation: 75.87 ± 16.95 mm2vs 132.45 ± 20.94 mm2, respectively; p < 0.001).ConclusionAlthough several quantitative brain MR markers provided high diagnostic accuracy in differentiating Progressive Supranuclear Palsy-Richardson's Syndrome from Parkinson's disease, the morphometric assessment of midbrain area is the best single diagnostic marker and should be routinely included in the neuroradiological work-up of parkinsonian patients.

Analysis of Uncertainties in Estimates of Components of Variance in Multivariate ROC Analysis

Rationale and Objectives Solutions have previously been presented to the problem of estimating the components of variance in the general linear model used for multivariate receiver operating characteristic (ROC) analysis. The case where the variance components do not change across the modalities under comparison was first treated, followed by the case where they are permitted to change. No analysis of uncertainties in these estimates has been presented previously. Materials and Methods For the case where the variance components do not change across modalities, the “jackknife-after-bootstrap” resampling procedure can be used together with conventional linear propagation of variance to solve for the uncertainties in estimates of the components. For the case where the components are permitted to change across modalities, a slight elaboration of this procedure is presented. Results The approach was validated by Monte Carlo simulations, where uncertainties in estimates of the variance components calculated by the jackknife-after-bootstrap procedure were found to converge in the mean to the Monte Carlo results over many independent trials. The method is exemplified with data from a study of readers—with and without the aid of a computer-assist modality—given the task of discriminating benign from malignant masses in mammography. Conclusion The present approach is relevant to a broad class of problems where estimates of multiple contributions to the variance observed in ROC assessment of diagnostic modalities are desired, in particular, for the assessment of multiple-reader studies of computer-aided diagnosis in radiology where the variance components may change across reading modalities (eg, unaided vs computer-aided reading).

Components-of-Variance Models for Random-Effects ROC Analysis: The Case of Unequal Variance Structures Across Modalities

Rationale and Objectives Several of the authors have previously published an analysis of multiple sources of uncertainty in the receiver operating characteristic (ROC) assessment and comparison of diagnostic modalities. The analysis assumed that the components of variance were the same for the modalities under comparison. The purpose of the present work is to obtain a generalization that does not require that assumption. Materials and Methods The generalization is achieved by splitting three of the six components of variance in the previous model into modality-dependent contributions. Two distinct formulations of this approach can be obtained from alternative choices of the three components to be split; however, a one-to-one relationship exists between the magnitudes of the components estimated from these two formulations. Results The method is applied to a study of multiple readers, with and without the aid of a computer-assist modality, performing the task of discriminating between benign and malignant clusters of microcalcifications. Analysis according to the first method of splitting shows large decreases in the reader and reader-by-case components of variance when the computer assist is used by the readers. Analysis in terms of the alternative splitting shows large decreases in the corresponding modality-interaction components. Conclusion A solution to the problem of multivariate ROC analysis without the assumption of equal variance structure across modalities has been provided. Alternative formulations lead to consistent results related by a one-to-one mapping. A surprising result is that estimates of confidence intervals and numbers of cases and readers required for a specified confidence interval remain the same in the more general model as in the restricted model.