Processing and analysis of serum antibody binding signals from Printed Glycan Arrays for diagnostic and prognostic applications

Our plain-English summary of Processing and analysis of serum antibody binding signals from Printed Glycan Arrays for diagnostic and prognostic applications, by M.I. Vuskovic, H. Xu, N.V. Bovin, et al (International Journal of Bioinformatics Research and Applications, 2011).

The study

• Authors: M.I. Vuskovic, H. Xu, N.V. Bovin, H.I. Pass, M.E. Huflejt
• Journal: International Journal of Bioinformatics Research and Applications
• Published: 2011;7(4):402-26
• Cited by: 20 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

Procedures for data preprocessing, quality control, data analysis, evaluation and visualization of the new high-throughput biomarker platform based on printed glycan arrays (PGA) are presented in this paper. PGAs are similar in concept to DNA arrays but contain deposits of various carbohydrate structures (glycans) instead of spotted DNAs. PGA biomarker discovery for the early detection, diagnosis and prognosis of human malignancies and viral diseases is based on the response of the immune system as measured by the level of binding of anti-glycan antibodies from human serum to the glycans on the array. Procedures related to PGA data processing are herein demonstrated in a pilot study of cases representing 50 sera from patients with malignant mesothelioma and a control sample of 65 sera from high risk subjects exposed to asbestos without symptoms of disease.

In plain English

This paper presented computer methods for checking, cleaning and analysing data from printed glycan arrays. These are glass slides carrying 211 sugar structures (glycans) that measure how strongly antibodies in a blood serum sample bind to each one. The authors demonstrated the methods on archived sera from 50 patients with malignant mesothelioma and 65 high-risk people exposed to asbestos who had no symptoms of disease. Fiber types were not described. Combining five glycans, the model correctly classified 74.1% of subjects under cross-validation, with an AUC (a 0-to-1 measure of how well groups are separated) of 0.811. On the data used to build it, the figures were 79.1% and 0.864. For prognosis, 35 patients were grouped as poor prognosis (27, with survival of 28 months or less) or good prognosis (8). Four of the 35 were still alive when clinical data were collected, so their survival times were incomplete. Two glycans gave a cross-validated accuracy of 85.7% (AUC 0.842). The authors concluded that the data showed diagnostic and prognostic power. They noted that the study was small and that larger, randomised sample sets are needed for potential clinical applications.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Methods (bioinformatics) paper demonstrated on a small pilot case-control data set of archived sera
• Who was studied: Archived sera from 50 patients diagnosed with malignant mesothelioma (cases) and 65 high-risk asbestos-exposed subjects without symptoms of disease (controls), from the NYU School of Medicine; 35 of the 50 patients were used for the survival analysis
• Exposure: Controls were described only as asbestos-exposed and high risk; no fiber type, job or dose was given. The measurement studied was serum antibody (IgG, IgM, IgA) binding to 211 glycans on printed glycan arrays
• Main results:
  • Each glycan alone separated mesothelioma patients from asbestos-exposed controls only modestly. The top-ranked glycan (GID 311) had p = 0.00003 and AUC 0.7274
  • With five glycans combined by multiple logistic regression (mesothelioma vs controls), cross-validated accuracy was 74.1% and AUC 0.811. Training (observed) values were 79.1% and 0.864
  • At the corrected decision line on the training set risk-score diagram, accuracy was 82.6%, specificity 92.3% and sensitivity 70%
  • A model including interaction terms (adding glycan GID 121) reached a maximal cross-validated AUC of 0.77, with an observed training AUC of 0.918
  • A Cox proportional hazards classifier for survival performed poorly, with accuracy below 60% and AUC below 0.6
  • Of the 35 mesothelioma patients in the survival analysis, four were still alive at the time of clinical data collection, and their survival times were treated as censored
  • Among 35 mesothelioma patients (27 labelled poor prognosis, with 28 months or less of survival, and 8 labelled good prognosis), two glycans (GID 154 and 215) gave a cross-validated accuracy of 85.7% and AUC of 0.842
  • Patients split into two groups by risk score on the training set had survival curves that differed (logrank p = 0.0049), with median survival times of 14 and 63 months. With leave-one-out cross-validation, the logrank p-value was 0.0112
• The authors' conclusion: The authors concluded that, despite small samples, they were able to show diagnostic and prognostic power of the glycan array data and their analytical approaches, and that the array-based predictors can be used for prognosis as well as diagnosis
• Limitations the authors state:
  • Small study; larger and randomised sample sets are needed for potential clinical applications
  • Cross-validated results for the interaction model were low because the samples were too small
  • The survival training sample was small and of limited use for identifying prognostic markers
  • 15 patients with very advanced, often unresectable cancer were excluded from the prognostic analysis
  • The proportionality assumption of the Cox model cannot be guaranteed for these predictors

Topics

Medical terms in this study, explained in our Lexicon: Biomarker, Discovery, Prognosis, Malignant Mesothelioma.

Related studies

• About the diagnostic value of BAP-1 antibody in malignant pleural mesothelioma: a meta-analysis — M. Mlika, M. Zorgati, et al. (2019)
• Evidence-based guidelines to optimize the selection of antibody panels in cytopathology: pleural effusions with malignant epithelioid cells — D.E. Westfall, X. Fan, et al. (2010)
• A diagnostic assay based on microRNA expression accurately identifies malignant pleural mesothelioma — H. Benjamin, D. Lebanony, et al. (2010)
• Serum levels of pantropic p53 protein and EGF-receptor, and detection of anti-p53 antibodies in former uranium miners (SDAG Wismut) — J. Schneider, P. Presek, et al. (1999)
• Tissue and serum EGFR as prognostic factors in malignant pleural mesothelioma — R. Gaafar, A. Bahnassy, et al. (2010)
• Serum concentrations of Krebs von den Lungen-6, surfactant protein D, and matrix metalloproteinase-2 as diagnostic biomarkers in patients with asbestosis and silicosis: a case-control study — C. Xue, N. Wu, et al. (2017)

Browse return to medical research.