Current progress in reverse vaccinology towards in silico vaccine discovery for parasitic diseases

Abstract text

Introduction

Reverse vaccinology (RV) was first described in 2000 as an in silico process that starts from the genomic sequence of the pathogen and ends with a list of potential vaccine candidates. Subsequently, other in silico related processes with overlapping workflows have also emerged with terms such as subtractive proteomics, computational vaccinology, and immunoinformatics. From the perspective of a new RV practitioner, determining the appropriate workflow steps and bioinformatics tools is a time consuming and overwhelming task.

Objectives

We document the current knowledge on RV and its usage in the research community by a comprehensive survey of scientific papers published with "reverse vaccinology" (RV) in the title from 2015 to 2021.

Materials & methods

All papers published with "reverse vaccinology" in their title were manually reviewed (140 papers in total, source: Web of Science). There were, however, 205 additional papers from the same period with RV in the abstract but not in the title. We made no attempt to capture current RV status in an unknown number of papers using an in silico vaccine discovery approach but with no reference to RV in the title or abstract.

Results

287 different bioinformatics tools were used in one or more of the RV workflows and some are more popular than others. 95.6% of the workflows rely on online tools. The RV pipelines developed so far mainly predict proteins naturally exposed to the immune system.

The workflow for selecting candidates in 87.8% of the latest publications involves a consecutive filtering process and not machine learning (ML). To automate the RV process, software pipelines were developed and made freely available since 2006. There are currently 11 known RV-related pipelines and almost all of them perform candidate ranking by ML for this purpose. They are not frequently used.

Currently, only 12.2% of the latest publications report tests on animal models providing a measure of success on whether an RV-derived candidate induces a protective response.

Conclusion

The majority of surveyed publications use classical RV as only one stage in a broader series of workflow stages to computationally identify vaccine candidates. Installation of a standalone program and/or adapt an API is a major disincentive to RV practitioners. We recommend "in silico vaccine discovery" should be consistently used in titles, abstracts, and/or keywords of future publications to unify the discipline area of RV.