Xuefeng B. Ling

Molecular classification of cancer types from microarray data using the combination of genetic algorithms and support vector machines

Simultaneous multiclass classification of tumor types is essential for future clinical implementations of microarray‐based cancer diagnosis. In this study, we have combined genetic algorithms (GAs) and all paired support vector machines (SVMs) for multiclass cancer identification. The predictive features have been selected through iterative SVMs/GAs, and recursive feature elimination post‐processing steps, leading to a very compact cancer‐related predictive gene set. Leave‐one‐out cross‐validations yielded accuracies of 87.93% for the eight‐class and 85.19% for the fourteen‐class cancer classifications, outperforming the results derived from previously published methods.

Multiclass cancer classification and biomarker discovery using GA-based algorithms

MOTIVATION The development of microarray-based high-throughput gene profiling has led to the hope that this technology could provide an efficient and accurate means of diagnosing and classifying tumors, as well as predicting prognoses and effective treatments. However, the large amount of data generated by microarrays requires effective reduction of discriminant gene features into reliable sets of tumor biomarkers for such multiclass tumor discrimination. The availability of reliable sets of biomarkers, especially serum biomarkers, should have a major impact on our understanding and treatment of cancer. RESULTS We have combined genetic algorithm (GA) and all paired (AP) support vector machine (SVM) methods for multiclass cancer categorization. Predictive features can be automatically determined through iterative GA/SVM, leading to very compact sets of non-redundant cancer-relevant genes with the best classification performance reported to date. Interestingly, these different classifier sets harbor only modest overlapping gene features but have similar levels of accuracy in leave-one-out cross-validations (LOOCV). Further characterization of these optimal tumor discriminant features, including the use of nearest shrunken centroids (NSC), analysis of annotations and literature text mining, reveals previously unappreciated tumor subclasses and a series of genes that could be used as cancer biomarkers. With this approach, we believe that microarray-based multiclass molecular analysis can be an effective tool for cancer biomarker discovery and subsequent molecular cancer diagnosis.

Integrative Urinary Peptidomics in Renal Transplantation Identifies Biomarkers for Acute Rejection

Noninvasive methods to diagnose rejection of renal allografts are unavailable. Mass spectrometry followed by multiple-reaction monitoring provides a unique approach to identify disease-specific urine peptide biomarkers. Here, we performed urine peptidomic analysis of 70 unique samples from 50 renal transplant patients and 20 controls (n = 20), identifying a specific panel of 40 peptides for acute rejection (AR). Peptide sequencing revealed suggestive mechanisms of graft injury with roles for proteolytic degradation of uromodulin (UMOD) and several collagens, including COL1A2 and COL3A1. The 40-peptide panel discriminated AR in training (n = 46) and test (n = 24) sets (area under ROC curve >0.96). Integrative analysis of transcriptional signals from paired renal transplant biopsies, matched with the urine samples, revealed coordinated transcriptional changes for the corresponding genes in addition to dysregulation of extracellular matrix proteins in AR (MMP-7, SERPING1, and TIMP1). Quantitative PCR on an independent set of 34 transplant biopsies with and without AR validated coordinated changes in expression for the corresponding genes in rejection tissue. A six-gene biomarker panel (COL1A2, COL3A1, UMOD, MMP-7, SERPING1, TIMP1) classified AR with high specificity and sensitivity (area under ROC curve = 0.98). These data suggest that changes in collagen remodeling characterize AR and that detection of the corresponding proteolytic degradation products in urine provides a noninvasive diagnostic approach.

AKI in hospitalized children: epidemiology and clinical associations in a national cohort.

BACKGROUND AND OBJECTIVES Although AKI is common among hospitalized children, comprehensive epidemiologic data are lacking. This study characterizes pediatric AKI across the United States and identifies AKI risk factors using high-content/high-throughput analytic techniques. DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS For the cross-sectional analysis of the 2009 Kids Inpatient Database, AKI events were identified using International Classification of Diseases, Ninth Revision, Clinical Modification codes. Demographics, incident rates, and outcome data were analyzed and reported for the entire AKI cohort as well as AKI subsets. Statistical learning methods were applied to the highly imbalanced dataset to derive AKI-related risk factors. RESULTS Of 2,644,263 children, 10,322 children developed AKI (3.9/1000 admissions). Although 19% of the AKI cohort was ≤ 1 month old, the highest incidence was seen in children 15-18 years old (6.6/1000 admissions); 49% of the AKI cohort was white, but AKI incidence was higher among African Americans (4.5 versus 3.8/1000 admissions). In-hospital mortality among patients with AKI was 15.3% but higher among children ≤ 1 month old (31.3% versus 10.1%, P<0.001) and children requiring critical care (32.8% versus 9.4%, P<0.001) or dialysis (27.1% versus 14.2%, P<0.001). Shock (odds ratio, 2.15; 95% confidence interval, 1.95 to 2.36), septicemia (odds ratio, 1.37; 95% confidence interval, 1.32 to 1.43), intubation/mechanical ventilation (odds ratio, 1.2; 95% confidence interval, 1.16 to 1.25), circulatory disease (odds ratio, 1.47; 95% confidence interval, 1.32 to 1.65), cardiac congenital anomalies (odds ratio, 1.2; 95% confidence interval, 1.13 to 1.23), and extracorporeal support (odds ratio, 2.58; 95% confidence interval, 2.04 to 3.26) were associated with AKI. CONCLUSIONS AKI occurs in 3.9/1000 at-risk US pediatric hospitalizations. Mortality is highest among neonates and children requiring critical care or dialysis. Identified risk factors suggest that AKI occurs in association with systemic/multiorgan disease more commonly than primary renal disease.

Significance Analysis and Multiple Pharmacophore Models for Differentiating P-Glycoprotein Substrates

P-glycoprotein (Pgp) mediated drug efflux affects the absorption, distribution, and clearance of a broad structural variety of drugs. Early assessment of the potential of compounds to interact with Pgp can aid in the selection and optimization of drug candidates. To differentiate nonsubstrates from substrates of Pgp, a robust predictive pharmacophore model was targeted in a supervised analysis of three-dimensional (3D) pharmacophores from 163 published compounds. A comprehensive set of pharmacophores has been generated from conformers of whole molecules of both substrates and nonsubstrates of P-glycoprotein. Four-point 3D pharmacophores were employed to increase the amount of shape information and resolution, including the ability to distinguish chirality. A novel algorithm of the pharmacophore-specific t-statistic was applied to the actual structure-activity data and 400 sets of artificial data (sampled by decorrelating the structure and Pgp efflux activity). The optimal size of the significant pharmacophore set was determined through this analysis. A simple classification tree using nine distinct pharmacophores was constructed to distinguish nonsubstrates from substrates of Pgp. An overall accuracy of 87.7% was achieved for the training set and 87.6% for the external independent test set. Furthermore, each of nine pharmacophores can be independently utilized as an accurate marker for potential Pgp substrates.

Urine Protein Biomarkers for the Diagnosis and Prognosis of Necrotizing Enterocolitis in Infants

OBJECTIVES To test the hypothesis that an exploratory proteomics analysis of urine proteins with subsequent development of validated urine biomarker panels would produce molecular classifiers for both the diagnosis and prognosis of infants with necrotizing enterocolitis (NEC). STUDY DESIGN Urine samples were collected from 119 premature infants (85 NEC, 17 sepsis, 17 control) at the time of initial clinical concern for disease. The urine from 59 infants was used for candidate biomarker discovery by liquid chromatography/mass spectrometry. The remaining 60 samples were subject to enzyme-linked immunosorbent assay for quantitative biomarker validation. RESULTS A panel of 7 biomarkers (alpha-2-macroglobulin-like protein 1, cluster of differentiation protein 14, cystatin 3, fibrinogen alpha chain, pigment epithelium-derived factor, retinol binding protein 4, and vasolin) was identified by liquid chromatography/mass spectrometry and subsequently validated by enzyme-linked immunosorbent assay. These proteins were consistently found to be either up- or down-regulated depending on the presence, absence, or severity of disease. Biomarker panel validation resulted in a receiver-operator characteristic area under the curve of 98.2% for NEC vs sepsis and an area under the curve of 98.4% for medical NEC vs surgical NEC. CONCLUSIONS We identified 7 urine proteins capable of providing highly accurate diagnostic and prognostic information for infants with suspected NEC. This work represents a novel approach to improving the efficiency with which we diagnose early NEC and identify those at risk for developing severe, or surgical, disease.

Urine Peptidomics for Clinical Biomarker Discovery

Urine-based proteomic profiling is a novel approach that may result in the discovery of noninvasive biomarkers for diagnosing patients with different diseases, with the aim to ultimately improve clinical outcomes. Given new and emerging analytical technologies and data mining algorithms, the urine peptidome has become a rich resource to uncover naturally occurring peptide biomarkers for both systemic and renal diseases. However, significant analytical hurdles remain in sample collection and storage, experimental design, data analysis, and statistical inference. This study summarizes, focusing on our experiences and perspectives, the progress in addressing these challenges to enable high-throughput urine peptidomics-based biomarker discovery.

Investigation of maternal environmental exposures in association with self-reported preterm birth

Identification of maternal environmental factors influencing preterm birth risks is important to understand the reasons for the increase in prematurity since 1990. Here, we utilized a health survey, the US National Health and Nutrition Examination Survey (NHANES) to search for personal environmental factors associated with preterm birth. 201 urine and blood markers of environmental factors, such as allergens, pollutants, and nutrients were assayed in mothers (range of N: 49-724) who answered questions about any children born preterm (delivery <37 weeks). We screened each of the 201 factors for association with any child born preterm adjusting by age, race/ethnicity, education, and household income. We attempted to verify the top finding, urinary bisphenol A, in an independent study of pregnant women attending Lucile Packard Childrens Hospital. We conclude that the association between maternal urinary levels of bisphenol A and preterm birth should be evaluated in a larger epidemiological investigation.

A diagnostic algorithm combining clinical and molecular data distinguishes Kawasaki disease from other febrile illnesses

BackgroundKawasaki disease is an acute vasculitis of infants and young children that is recognized through a constellation of clinical signs that can mimic other benign conditions of childhood. The etiology remains unknown and there is no specific laboratory-based test to identify patients with Kawasaki disease. Treatment to prevent the complication of coronary artery aneurysms is most effective if administered early in the course of the illness. We sought to develop a diagnostic algorithm to help clinicians distinguish Kawasaki disease patients from febrile controls to allow timely initiation of treatment.MethodsUrine peptidome profiling and whole blood cell type-specific gene expression analyses were integrated with clinical multivariate analysis to improve differentiation of Kawasaki disease subjects from febrile controls.ResultsComparative analyses of multidimensional protein identification using 23 pooled Kawasaki disease and 23 pooled febrile control urine peptide samples revealed 139 candidate markers, of which 13 were confirmed (area under the receiver operating characteristic curve (ROC AUC 0.919)) in an independent cohort of 30 Kawasaki disease and 30 febrile control urine peptidomes. Cell type-specific analysis of microarrays (csSAM) on 26 Kawasaki disease and 13 febrile control whole blood samples revealed a 32-lymphocyte-specific-gene panel (ROC AUC 0.969). The integration of the urine/blood based biomarker panels and a multivariate analysis of 7 clinical parameters (ROC AUC 0.803) effectively stratified 441 Kawasaki disease and 342 febrile control subjects to diagnose Kawasaki disease.ConclusionsA hybrid approach using a multi-step diagnostic algorithm integrating both clinical and molecular findings was successful in differentiating children with acute Kawasaki disease from febrile controls.

Urine Peptidomic and Targeted Plasma Protein Analyses in the Diagnosis and Monitoring of Systemic Juvenile Idiopathic Arthritis

PurposeSystemic juvenile idiopathic arthritis is a chronic pediatric disease. The initial clinical presentation can mimic other pediatric inflammatory conditions, which often leads to significant delays in diagnosis and appropriate therapy. SJIA biomarker development is an unmet diagnostic/prognostic need to prevent disease complications.Experimental DesignWe profiled the urine peptidome to analyze a set of 102 urine samples, from patients with SJIA, Kawasaki disease (KD), febrile illnesses (FI), and healthy controls. A set of 91 plasma samples, from SJIA flare and quiescent patients, were profiled using a customized antibody array against 43 proteins known to be involved in inflammatory and protein catabolic processes.ResultsWe identified a 17-urine-peptide biomarker panel that could effectively discriminate SJIA patients at active, quiescent, and remission disease states, and patients with active SJIA from confounding conditions including KD and FI. Targeted sequencing of these peptides revealed that they fall into several tight clusters from seven different proteins, suggesting disease-specific proteolytic activities. The antibody array plasma profiling identified an SJIA plasma flare signature consisting of tissue inhibitor of metalloproteinase-1 (TIMP1), interleukin (IL)-18, regulated upon activation, normal T cell expressed and secreted (RANTES), P-Selectin, MMP9, and L-Selectin.Conclusions and Clinical RelevanceThe urine peptidomic and plasma protein analyses have the potential to improve SJIA care and suggest that SJIA urine peptide biomarkers may be an outcome of inflammation-driven effects on catabolic pathways operating at multiple sites.