Abstract
Detection of protection-associated epitopes via reverse vaccinology is the first step for development of subunit vaccines against microbial pathogens. Mapping subunit vaccine targets requires high throughput methods, which would allow delineation of epitopes recognized by protective antibodies on a large scale. Phage displayed random peptide library coupled to Next Generation Sequencing (PDRPL/NGS) is the universal platform that enables high-yield identification of peptides that mimic epitopes (mimotopes). Despite being unsurpassed as a tool for discovery of polyclonal serum mimotopes, the PDRPL/NGS is far inferior as a quantitative method of immune response. Difficult-to-control fluctuations in amounts of antibody-bound phages after rounds of selection and amplification diminish the quantitative capacity of the PDRPL/NGS. In an attempt to improve the accuracy of the PDRPL/NGS method, we compared the discriminating capacity of two approaches for PDRPL/NGS data analysis. The whole-unique-sequence-based analysis (WUSA) involved generation of 7-mer peptide profiles and comparison of the numbers of sequencing reads for unique peptide sequences between serum samples. The motif-based analysis (MA) included identification of 4-mer consensus motifs unifying unique 7-mer sequences and comparison of motifs between serum samples. The motif comparison was based not on the numbers of sequencing reads, but on the numbers of distinct 7-mers constituting the motifs. Our PDRPL/NGS datasets generated from biopanning of protective and non-protective anti-Borrelia burgdorferi sera of New Zealand rabbits were used to contrast the two approaches. As a result, the principle component analyses (PCA) showed that the discriminating powers of the WUSA and MA were similar. In contrast, the unsupervised hierarchical clustering obtained via the MA classified the preimmune, non-protective, and protective sera better than the WUSA-based clustering. Also, a total number of discriminating motifs was higher than that of discriminating 7-mers. In sum, our results indicate that MA approach improves the accuracy and quantitative capacity of the PDRPL/NGS method.