Chapter 4
Computational Antigen Discovery for Eukaryotic Pathogens
Using Vacceed
Stephen J. Goodswen, Paul J. Kennedy, and John T. Ellis
Abstract
Bioinformatics programs have been developed that exploit informative signals encoded within protein
sequences to predict protein characteristics. Unfortunately, there is no program as yet that can predict
whether a protein will induce a protective immune response to a pathogen. Nonetheless, predicting those
pathogen proteins most likely from those least likely to induce an immune response is feasible when
collectively using predicted protein characteristics. Vacceed is a computational pipeline that manages
different standalone bioinformatics programs to predict various protein characteristics, which offer supporting evidence on whether a protein is secreted or membrane -associated. A set of machine learning
algorithms predicts the most likely pathogen proteins to induce an immune response given the supporting
evidence. This chapter provides step by step descriptions of how to configure and operate Vacceed for a
eukaryotic pathogen of the user’s choice.
Key words Vacceed, Machine learning, In silico vaccine discovery, Computational antigen discovery,
Eukaryotic pathogen
1 Introduction
Protein sequences are not random assemblies of amino acids. There
is a precise biological reason why one particular amino acid is
connected to another, which ultimately contributes to a protein’s
distinctive characteristics [1]. Researchers, over the last two decades, have developed bioinformatics programs that exploit informative signals or patterns encoded within these amino acid
sequences to predict protein characteristics. Examples of these
characteristics are subcellular localization [2], presence and location
of signal peptide cleavage sites [3], and transmembrane topology
[4]. With respect to discovering protein vaccine candidates, no
signal has yet been detected that helps predict a characteristic
signifying a protein’s contributing capacity to a protective immune
response in a host. Consequently, the current computational antigen discovery aspiration is to distinguish those pathogen proteins
Blaine A. Pfeifer and Andrew Hill (eds.), Vaccine Delivery Technology: Methods and Protocols, Methods in Molecular Biology,
vol. 2183, https://doi.org/10.1007/978-1-0716-0795-4_4, © Springer Science+Business Media, LLC, part of Springer Nature 2021
29
Computational Antigen Discovery for Eukaryotic Pathogens
Using Vacceed
Stephen J. Goodswen, Paul J. Kennedy, and John T. Ellis
Abstract
Bioinformatics programs have been developed that exploit informative signals encoded within protein
sequences to predict protein characteristics. Unfortunately, there is no program as yet that can predict
whether a protein will induce a protective immune response to a pathogen. Nonetheless, predicting those
pathogen proteins most likely from those least likely to induce an immune response is feasible when
collectively using predicted protein characteristics. Vacceed is a computational pipeline that manages
different standalone bioinformatics programs to predict various protein characteristics, which offer supporting evidence on whether a protein is secreted or membrane -associated. A set of machine learning
algorithms predicts the most likely pathogen proteins to induce an immune response given the supporting
evidence. This chapter provides step by step descriptions of how to configure and operate Vacceed for a
eukaryotic pathogen of the user’s choice.
Key words Vacceed, Machine learning, In silico vaccine discovery, Computational antigen discovery,
Eukaryotic pathogen
1 Introduction
Protein sequences are not random assemblies of amino acids. There
is a precise biological reason why one particular amino acid is
connected to another, which ultimately contributes to a protein’s
distinctive characteristics [1]. Researchers, over the last two decades, have developed bioinformatics programs that exploit informative signals or patterns encoded within these amino acid
sequences to predict protein characteristics. Examples of these
characteristics are subcellular localization [2], presence and location
of signal peptide cleavage sites [3], and transmembrane topology
[4]. With respect to discovering protein vaccine candidates, no
signal has yet been detected that helps predict a characteristic
signifying a protein’s contributing capacity to a protective immune
response in a host. Consequently, the current computational antigen discovery aspiration is to distinguish those pathogen proteins
Blaine A. Pfeifer and Andrew Hill (eds.), Vaccine Delivery Technology: Methods and Protocols, Methods in Molecular Biology,
vol. 2183, https://doi.org/10.1007/978-1-0716-0795-4_4, © Springer Science+Business Media, LLC, part of Springer Nature 2021
29