Bayesian Hierarchical Modeling for Subject-Level Response Classification in Peptide Microarray Immunoassays

Abstract

Summary The peptide microarray immunoassay simultaneously screens sample serum against thousands of peptides, determining the presence of antibodies bound to array probes. Peptide microarrays tiling immunogenic regions of pathogens (e.g. envelope proteins of a virus) are an important high throughput tool for querying and mapping antibody binding. Because of the assay’s many steps, from probe synthesis to incubation, peptide microarray data can be noisy with extreme outliers. In addition, subjects may produce different antibody profiles in response to an identical vaccine stimulus or infection, due to variability among subjects’ immune systems. We present a robust Bayesian hierarchical model for peptide microarray experiments, pepBayes, to estimate the probability of antibody response for each subject/peptide combination. Heavy-tailed error distributions accommodate outliers and extreme responses, and tailored random effect terms automatically incorporate technical effects prevalent in the assay. We apply our model to two vaccine trial datasets to demonstrate model performance. Our approach enjoys high sensitivity and specificity when detecting vaccine induced antibody responses. A simulation study shows an adaptive thresholding classification method has appropriate false discovery rate control with high sensitivity, and receiver operating characteristics generated on vaccine trial data suggest that pepBayes clearly separates responses from non-responses.

Co-immunization of DNA and Protein in the Same Anatomical Sites Induces Superior Protective Immune Responses against SHIV Challenge

“…The following references appear in the Supplemental Information: He and Fong (2019); Huang et al (2009); and Imholte and Gottardo (2016).…”

“…Absorbance at 450nm was measured and standard curves were generated using SoftMax Pro (Molecular Devices) to calculate total IgG levels in samples. Linear peptide array binding assay as previously described (Imholte and Gottardo, 2016;Karasawas et al, 2012). Microarray binding was performed using the HS4800 Pro Hybridization Station (Tecan, Mä nnedorf, Switzerland) and scanned using an Axon Genepix 4300 Scanner (Molecular Devices, Sunnyvale, CA, USA).…”

EPIphany—A Platform for Analysis and Visualization of Peptide Immunoarray Data

“…Several papers describe immunoarray results obtained in the absence of data normalization that have been validated through independent techniques, and arguments have been presented against the need for any normalization ( Hecker et al, 2012 ). Other reports have demonstrated added value in applying various normalization methods (i.e., linear model and Bayesian hierarchical modeling) and data pre-processing methods (i.e., correcting for spatial and systematic biases) for discriminating binding signatures between cohorts in a vaccine study ( Renard et al, 2011 ; Imholte et al, 2016 ; Imholte and Gottardo, 2016 ). At the present time, we feel it is premature to discount the value and appropriateness of data pre-processing and normalization methods given there is no consensus among the current end-users.…”

“…Several public web applications analyze user-submitted immunoarray data to identify binding motifs and profiles, but do not compare binding signatures across cohorts, including ArrayPitope ( Hansen et al, 2017 ) and SVM-PEPARRAY ( Chen et al, 2009 ). Additionally, tools such as rapMad ( Renard et al, 2011 ), pepStat ( Imholte et al, 2016 ), and pepBayes ( Imholte and Gottardo, 2016 ) are available to compare binding signatures from different cohorts, but only exist as R packages. Collectively these resources face several critical limitations: requirement of computational biology expertise outside the realms of many biological researchers, the absence of a web-based interface, the need to be run locally by the user, insufficient resources for statistical interpretation, an absence of data normalization options, and a lack of tools for visualization of the results.…”

Ranking Antibody Binding Epitopes and Proteins Across Samples from Whole Proteome Tiled Linear Peptides

“…There are many existing methods for analyzing microarray and peptide array data [2][3][4][5][6][7][8][9][10][11][12][13][14][15], including pepStat, which was designed to analyze different viral strains for a single protein, and a few methods for analyzing these ultra-dense, highdimensional array data [13,15,16], but additional methods are necessary to advance understanding of immune response as antibody binding signals from arrays represent an aggregate of complex biological and biochemical interactions. Antibodies may bind to peptides via various mechanisms of interaction based on the antibody's antigen binding site and amino acid configuration of the peptides, affecting binding affinity, and each protein may have multiple epitopes that are bound by antibodies, amplifying the immune response to the protein.…”