Vaccine Target Discovery

Abstract

With the rise in novel infectious agents and disease pandemics, a new era of vaccine discovery is necessary. To address this, the new field of immunomics is described, which is synergistically powered by integrating bioinformatics methodologies with technological advances in biology and high-throughput instrumentation. By incorporating biological data from immunology and molecular biology with current genomics and proteomics, immunomics is geared to deliver an insight into immune function, optimal stimulation of immune responses and precise mapping and rational selection of immune targets that cover antigenic diversity. These efforts are expected to contribute towards the development of new generation of vaccines, tailored to both the genetic make-up of the human population and of the pathogen. Vaccine technologies are also being explored for prevention or control of non-communicable diseases.

Introduction

Vaccination is one of the most efficacious medical interventions that have decreased human morbidity and mortality in all regions of the world. It not only has dramatically reduced the incidence of numerous diseases (Hilleman, 1985, Andre et al., 2008), such as measles, diphteria, mumps, rubella, tetanus, yellow fever, pertussis, and poliomyelitis, but also eradicated the dreaded smallpox viral infection (Strassburg, 1980). Efforts are currently also focused on developing vaccines for treatment of other diseases (Ada, 2003). This triumph over infectious diseases has been achieved by using either killed or attenuated conventional vaccines. A conventional or prophylactic vaccination is based on deliberate exposure to non-virulent form of the pathogen to establish immunity from subsequent exposure to the virulent form of the pathogen (Lee et al., 2012). This approach requires little knowledge of the molecular nature of the individual pathogen antigens or the immune responses they elicit. To date, this approach has undoubtedly been the most successful.

There is, however, an ongoing trend towards emerging infectious diseases (Fauci, 2001, Fauci et al., 2005) against humankind in different parts of the world. These diseases are caused by the emergence of new pathogens, resurgence of old ones, constantly mutating pathogens, drug-resistant pathogens, and even include pathogens used as agents of bioterrorism. Previously undescribed pathogens, such as severe acute respiratory syndrome (SARS), bird flu (avian influenza), mad cow, and HIV/AIDS, are appearing at an increasing frequency. On the other hand, old diseases such as ebola, hanta, dengue and cholera, among others, are invading populations from which they have disappeared or cross species barrier to invade new species. Occasionally, pathogens mutate (or recombine) and then they adapt and show up as new variants that evade the host’s acquired immunity. The annual influenza epidemics are generally due to genetically drifting strains of influenza that differ slightly from previous strains (Bouvier and Palese, 2008). The development of drug-resistance pathogens, such as malaria, pneumococci, enterococci, and tuberculosis, has increased over the years, partly due to the widespread and inappropriate use of drugs (Knobler et al., 2003). Moreover, the pathogens responsible for the emerging infectious diseases are also of potential use as bioterrorist weapon (Ryan, 2008). For these reasons, emerging infectious diseases continue to pose threats to public health (Morens and Fauci, 2013). Therefore, there is a need for more effective vaccines to help reduce human morbidity and mortality from emerging infectious diseases.

The last decade has seen significant advances in new technologies for the development of new vaccines. These technologies, combined with our understanding of host response to foreign antigens, have laid the foundation for rapid advances in vaccinology. Few potential candidate approaches for this new family of vaccines (Arnon and Ben-yedidia, 2003, Minichiello, 2002, Babiuk, 1999, Nandy and Basak, 2016) are subunit vaccines, genetically engineered live vaccines, and polynucleotide (DNA or genetic) vaccines. All these new directions in vaccine development have something in common; their most important challenge is the discovery of key antigens from the array of proteins encoded by the pathogen genome that are able to elicit protective immune response against these pathogens and are effective for majority of the human population. The presence of genetic variation in the genes of the host immune system across the human population and the genomes of the pathogen variants make this a multi-dimensional and a combinatorial problem.

Since 1980s, the focus of vaccine design has been on the pathogen’s variable domains, mutants, multivalent coverage and not so much on conserved regions (Plotkin, 2005). However, a large body of data is building up in the field of vaccine design and epitope prediction that points to the neglected aspect of conserved epitopes as important targets of vaccine development. In fact, evidence is accumulating that while many variable regions are highly antigenic, sufficiently large number of them are actually non-immunoprotective and have been exploited by viruses and other pathogens for immune escape and may lead to immunopathology (Haydon and Woolhouse, 1998). Conserved epitopes were thought to be un-important due to their lack of immunogenicity (Bona et al., 1998; Li et al., 2011). However, the immunogenicity of such conserved epitopes especially those which are immune-protective and conserved amongst many variants and mutant strains, can be boosted using adjuvants, which are chemicals or approved drug molecules. Moreover, conserved epitopes also help address the issues of ethnicity (“pan-haplotype responsive”) compared to variable epitopes (Khan et al., 2008). In addition, the conserved epitopes can circumvent vaccine design issues of escape mutants and the need of using the latest strains. In particular, if a set of conserved, immunogenic, and immunoprotective epitopes are suitably boosted, they will not elicit mutations in the pathogen, and thus can be reasonably predicted to remain unchanged in the next season of infection. The possibility of extended efficacy of vaccines would be a tremendous advance for the vaccine industry and will potentially serve the global population with protection against infections (Sylvester-Hvid et al., 2002, De Groot et al., 2004, Sette et al., 2001, Raman et al., 2014, Khan et al., 2017).

Vaccine informatics, a fledgling sub-field of reverse vaccinology, has the potential to develop effective vaccines (Hegde et al., 2017, Khan et al., 2017). With the rapid expansion of vaccine related data (host and pathogen) stemming from both classical and high-throughput genomic/proteomic approaches, identifying conserved, robust, immunogenic, and immunoprotective epitopes manually from this large data pool is inefficient. Vaccine informatics is a practical science for designing new vaccines with a focus on bioinformatics-driven acquisition, manipulation and analysis of data related to the immune system and disease agents (Raman et al., 2014). It provides a means for systematic study of big data, pre-screening of targets, and facilitates experimental design for validation by a small number of key experiments. The bioinformatics support can be divided into two, the standard bioinformatics support and the more specialised immunoinformatics support (Petrovsky et al., 2003). The standard support includes basic bioinformatics functions, such as sequence comparison and alignment, database searching, hunting for patterns and profiles, 3D-structure analysis and modeling, and data annotation (reviewed in Brusic and Petrovsky (2003)). Immunoinformatics is a more targeted bioinformatics support with an emphasis on data-warehousing and mining of immunological data, such as prediction of immunogenicity (Soria-Guerra et al., 2015, Brusic and Petrovsky, 2003). Vaccine researchers are taking advantage of these bioinformatics approaches, in combination with experimental validation, to discover and facilitate better understanding of the components of the human immunome, which then aid in the design of new vaccines. The immunome can be defined as the complete set of genes and proteins of the immune system. Highly accurate target predictions can diminish discovery cost by 10–20 folds (De Groot et al., 2002, Kast et al., 1994).

Background

In the human and higher vertebrate host, the major functions of the immune system are the maintenance of homeostasis, surveillance and tolerance to self-structures, and defence followed by immunity against pathogens (Yatim and Lakkis, 2015). The immune system is widely distributed in the body and comprises of immune organs, tissues, and cells, connected as a complex, but tightly regulated network (Jerne, 1993, NIH, 2003, Nicholson, 2016). In general, the processes that take place at the molecular level and cellular level largely initiate and regulate the function of the immune system. A healthy immune system will discriminate ‘non-self’ or foreign antigenic proteins from those that are normally present (‘self’) in an organism and will raise appropriate responses. The number of self-structures is large, but finite; the number of non-self structures is practically infinite. In the cell, both self and non-self proteins are digested by the proteasomes into short peptide fragments, which are then bound by major histocompatibility complex (MHC) molecules to form peptide/MHC complexes and are displayed on the surface of host cells (Vigneron and Van den Eynde, 2014). These peptides are recognition labels, which display the contents of host cells to T cells of the immune system. The presence of non-self peptides is a prerequisite for the initiation of immune responses. Peptides produced by degradation of intracellular proteins bind MHC class I molecules and are recognised by CD8⁺ T cell receptor (Shastri et al., 2002, Chowell et al., 2015, Blum et al., 2013). MHC class II molecules present peptides, produced by degradation of proteins of extracellular origin, on the surface of antigen-presenting cells to CD4⁺ T cells (reviewed in Lennon-Duménil et al. (2002)). A major function of CD8⁺ T cells is to recognize and destroy cells infected by pathogens (Nicholson, 2016). Peptides displayed by the MHC class II molecules mainly serve to regulate immune responses; they are crucial for the initiation, enhancement and suppression of immune responses.

Driven by methods of molecular and cell biology, significant advances in the understanding of immunological processes have been made during the last two decades. This progress over the years resulted in the continuous accumulation of huge amount of immunological data obtained experimentally. This growing number of immunological data and the high complexity of the functional and structural foundation of the immune processes created a need for improved data management to enable advance data analysis. The need to manage and analyse this growing amount of complex data has led to the development of a number of immunological databases (Brusic et al., 2000), such as SYFPEITHI and MCHPEP, IMGT and FIMM, and complex computational models (Petrovsky and Brusic, 2002). The purpose of immunological databases is to facilitate the collection of, access to, and use of immunologically relevant data. One of the applications of the databases in immunology is for vaccine research and development by using complex computational models in combination with experimental approaches to help precise our understanding of antigen presentation and recognition by the immune system.

The combined effort between experimental and computational immunology provided us with a new perspective to designing vaccines that will be effective across demographic boundaries. Previously, the extreme degree of polymorphism observed in the MHC posed limitations for the development of such a vaccine because the ability to trigger an effective T-cell response is partly determined by the MHC phenotype of the individual and different individuals have different MHC allele (Macdonald et al., 2001, Marrack et al., 2017). The MHC genes are the most polymorphic of all human genes, with more than 10,000 alleles known (as of Feb 2018), and are important in increasing the range of responses that different individuals can mount. In humans, this polymorphism results from concentrated amino acid substitutions in the peptide-binding groove of human leukocyte antigen (HLA, the human MHC system) molecules that produce variability in peptide binding and presentation to T cells (MacDonald et al., 2000). Over the years, few groups have investigated the possibility of a functional classification of HLA polymorphism based on peptide-binding specificities. It was found that majority of HLA alleles (both class I and II) could be grouped into 18 or more different ‘supertypes’, purely on the basis of similarities in their peptide binding specificity (Lund et al., 2004). It has been suggested that the majority of all major human populations can be covered by only few HLA supertypes, where the different members of each supertype bind similar peptides (‘promiscuous peptides’) for presentation to T-cell (Sette et al., 1999). Further, latest developments show evidence for presence of “immunological hot-spots” (Srinivasan et al., 2004) in antigens. Immunological hot-spots are defined as antigenic regions possessing multiple promiscuous peptides that are supertype specific. This area of research raises the prospect of identifying key pathogenic antigens that possess immunological hot-spots as best candidates for vaccine design as it will provide protection at the population level, irrespective of ethnicity.

The humoral response involves antibodies produced by B-cells, which recognize both linear and conformational B-cell epitopes on the surface of the pathogen. Conformational neutralizing epitopes are the primary focus of various vaccine research for protective humoral responses. However, unlike linear B-cell epitopes and T-cell epitopes, reliable computational tools for prediction of conformational epitopes are limited (Kulkarni-Kale et al., 2005, Zhang et al., 2011).

As for the pathogens, the last decade witnessed the rapid expansion of sequence data at our disposal, stemming from both genomic and proteomic approaches, enabling analysis to map key antigens that are potential targets for protective immune responses. The genomic sequences of a large number of pathogens that threaten public health have been completed or are impending completion. For example, the whole genome of over 7,475 viruses have been sequenced and deposited in the major public database Entrez Genome (see “Relevant Website section”). The data derived from the genome/proteome sequencing and related projects for a particular species of pathogen, such as West Nile Virus, is the missing gap towards the development of vaccines effective against the majority of the variants currently known within that pathogenic species and probably against novel ones that are yet to emerge (Koo et al., 2009). Such vaccines will be much superior to the current generation of vaccines, which are solely based on a single or few antigens providing protection only against certain variants of a pathogen and therefore might elicit too narrow a breadth of response to provide protection from the remaining diverse variants of the same pathogen species (Doolan, 2003).

We have made significant progress in understanding the processes in the host that are involved in mounting an immune response. However, we are still far from having a good understanding of the natural complexity of the pathogens. A good starting point would be by utilizing the pathogens genomic/proteomic data to study their sequence diversity, and identify antigens containing conserved and variable immunological hot-spots. To fully realize the promise of the available datasets for such study, we would require the development of appropriate technologies for systematically converting genomic/proteomic data into protective vaccines. Recently, systematic genome-wide approach to identify the key antigens of a pathogenic species from the numerous variant sequences of the same pathogen have been reported (Dhanda et al., 2017, Rizwan et al., 2017, Goodswen et al., 2014, Vivona et al., 2006, Del Tordello et al., 2016, María et al., 2017, Doolan, 2003, Raman et al., 2014, Koo et al., 2009, Khan et al., 2017). The selected key antigens will represent the minimal representative sets of target sequences required to provide immunity against the majority of the existing variants of a particular pathogen species.

Bioinformatics is an inter-disciplinary field that is essential for the analysis and interpretation of complex and large quantity of biological data generated by functional studies and high throughput technologies. It is used to propose the next sets of experiments and, most importantly, to derive better understanding of biological processes. As stated, the number of pathogen sequence data in public databases is increasing rapidly, however, experimental approaches to study this large data pool for the development of immune interventions are time-consuming, costly and almost impractical. Through combination of bioinformatics and experimental approaches, it is possible to select key experiments and help optimize experimental design. Computer algorithms are increasingly used to speed-up the process of knowledge discovery by helping to identify critical experiments for testing hypothesis built upon the result of computational screening. A number of successful examples for application of computer models to study immunological problems have been described in Brusic et al. (2005). Such examples illustrate the power of computational approach to complex problems involving potentially vast datasets with potential biases, errors and discrepancies.

A Computational Framework for Vaccine Target Discovery

Reverse vaccinology, a bottom-up genomic approach, has been successfully applied to the development of vaccines against pathogens that were previously not suited to such development (Vernikos, 2008, Rappuoli and Covacci, 2003, Rappuoli, 2001, Del Tordello et al., 2016). The pre-requisite for this approach is the sequence data of the target pathogen, which acts as input to various bioinformatics algorithms for prediction of putative antigens that are likely to be successful vaccine targets. These candidates can then be validated by a small number of key experiments in the lab. The approach has been successfully applied to the development of universal vaccines against group B Streptococcus (Maione et al., 2005) and vaccine candidates against MenB (Pizza et al., 2000), among others (Rappuoli and Covacci, 2003). Reverse vaccinology is a promising method for the high-throughput discovery of candidate vaccine targets that have the potential to mirror the dynamics and antigenic diversity of the target pathogen population, which includes the diversity of the interacting partner, the immune system. However, a big challenge to this end is the need to understand how vaccine developers can cover antigenic diversity and develop a systematic approach to rationally screen pathogen data to select candidate vaccine targets that cover the diversity.

Over the years, a number of bioinformatics pipelines have been designed that predict vaccine candidates both rapidly and efficiently. VacSol (Rizwan et al., 2017), NERVE (Vivona et al., 2006), and Vacceed (Goodswen et al., 2014) are examples of such pipelines for proteomes of bacterial or eukaryotic pathogens and these pipelines are highly configurable and scalable (Zaharieva et al., 2017). They include multiple steps and integrate various algorithms for analysis and comparison. Shortlisted candidate vaccine targets are ranked for prioritization towards experimental validation. These pipelines are expected to improve the vaccine target discovery process.

The general characteristics desired for a candidate vaccine target are (i) highly conserved; (ii) pathogen-specific; (iii) important for structure/function; (iv) immune-relevant; and (v) antigenically similar to circulating strains. Highly conserved targets are less likely to mutate and escape immune recognition. This is particularly so if they are important for structure and function, suggesting a robust historical conservation. High conservation also reduces the possibility of altered-peptide ligand (APL) effect from variant epitopes of the same pathogen species (Sloan-Lancaster and Allen, 1996, Evavold et al., 1993, Rothman, 2004). Variants may also originate from other pathogens, in particular those that co-circulate or co-infect with the pathogen of interest and if they belong to the same family. In vaccine design, epitopes common to other pathogens could either be useful by inducing cross-protection, or detrimental by inducing altered-ligand effect. Thus, potential vaccine targets should be analyzed for specificity to the target pathogen. The definition of virus species-specific vaccine targets can be further expanded to exclude those with one amino acid mismatch to human sequences in order to avoid possibility of molecular mimicry. Antigenic mismatch between a vaccine and circulating strains has been shown to increases the risk of disease outbreak by 1000-fold compared to immunization using identical strains (Park et al., 2004). Thus, it is important to assess the extent of antigenic identity between the candidate vaccine targets and the circulating strains.

Typically, a generic semi-automated computational framework comprises of three key components: data collection, data processing, and data analysis (Khan, 2005, Khan, 2009, Khan et al., 2006, Khan et al., 2017). Fig. 1 illustrates the workflow of the framework and Fig. 2 provides a non-comprehensive list of commonly used tools. Data collection would involve the user providing a set of sequences (aligned or unaligned) of the pathogen of interest for analysis. The sequences could be a single protein dataset, multiple or the complete proteome. Additionally, comparative analyses of the sequences can be performed between subtypes/groups of the pathogen. The sequences can be retrieved from public repositories, primary or specialist databases, such as the NCBI Entrez Protein database (NCBI Resource Coordinators, 2017) or Influenza Research Database (IRD) (Zhang et al., 2017), respectively. User can also provide sequences derived from their experimental work, not available in public databases. Typically, downloaded data would comprise full-length or partial sequences, and the corresponding metadata. Data processing would involve removal of duplicate sequences from the dataset, and for comparative analysis, the merging of the input sequences will be required prior to the alignment step. Multiple sequence alignment will be carried out using an existing tool that is robust in dealing with large sequence data, such as (Sievers et al., 2011; Katoh et al., 2017; Edgar, 2004; Do et al., 2005) Clustal Omega, MAFFT, MUSCLE or PROBCONS. The output alignment quality will be manually inspected for any errors and/or misalignments, which are common when dealing with partial sequences. Henceforth, the data would be ready for analyses, which can involve performing a diversity analysis, such as by measuring entropy values (Heiny, 2007, Hu et al., 2013, Khan, 2008, Koo et al., 2009) and quantifying variant motifs for each, user-defined, k-mer positions in the alignment. The results of these analyses will be plotted as an output for the user, providing a holistic view of the diversity, including variant distribution. The user can define a preferred conservation threshold for selection of highly conserved sequences. These selected sequences are then analyzed for distribution of variants in nature (Khan, 2008, Koo et al., 2009), including matches to human proteins, enabling the identification of pathogen specific, highly conserved sequences. The robust nature of the historical conservation can be assessed by performing functional and structural analysis (Sprenger, 2008, Hung and Link, 2011, Wizemann et al., 1999, Sachdeva et al., 2005, Monterrubio-López et al., 2015, He, 2014, Tang et al., 2014, María et al., 2017). The relevance of the identified potential candidate vaccine targets for use against current circulating strains can be assessed by measuring the incidence of the candidate targets in the corresponding sequences of recent strains of the virus of interest (Khan et al., 2017). The antigenicity/immunogenicity of the candidate sequences are predicted to assess their immune relevance. Experimental validation includes matching the predicted epitopes with reported epitopes in public databases, such as the Immune Epitope Database (IEBD) or SYFPEITHI, or performing quantitative measurements of the pre-selected candidate peptides by generating synthetic constructs and testing for their immunogenicity, such as by use of HLA transgenic animal models that express specific HLA alleles (Rosloniec et al., 1997, Lefranc et al., 2009, Gourlay et al., 2017, Khan, 2005).

Fig. 1.

Computational framework for vaccine target discovery.

Fig. 2.

An example of commonly used tools and databases for vaccine target discovery.

Immunoinformatics

Bioinformatics tools can facilitate the process of epitope mapping by identifying peptides that can potentially elicit T-cell responses. Binding of epitopes to HLA antigens is highly allele-specific; core peptide-binding motifs (usually between 8- and 11-mers, most often 9-mers) have been defined experimentally for a number of HLA class I and class II alleles and incorporated into computational algorithms, allowing to predict candidate HLA-binding epitopes in silico from protein sequences (Parker et al., 1994, Rammensee et al., 1999, Sturniolo et al., 1999, Zhang, 2005, Nielsen, 2010, Karosiene et al., 2013, Paul et al., 2013, Paul et al., 2015b, Paul et al., 2015a, Andreatta et al., 2015, Pro et al., 2015, Trolle et al., 2015, Jurtz, 2017, Abelin et al., 2017, Fleri et al., 2017). More recently, quantifiable predictive features of TCRαβ binding to HLA/epitope complexes have been also described (Birnbaum et al., 2014, Dash et al., 2017, Glanville et al., 2017, Gee et al., 2017).

Two main categories of specialized immunoinformatics tools are available for prediction of MHC binding peptides – methods based on identifying patterns in sequences of binding peptides, and those that employ three-dimensional (3D) structures to model peptide/MHC interactions (Tong et al., 2007, Liljeroos et al., 2015, Gourlay et al., 2017). Pattern-based methods includes binding motifs, quantitative matrices, decision trees, artificial neural networks (ANNs), hidden Markov models (HMMs) and support vector machines (SVMs), among others. In contrast, the structure based methods are theoretically rooted and include homology modeling, docking and 3D threading techniques (Dominguez et al., 2003, De Vries et al., 2010, Agostino et al., 2016, Khan and Ranganathan, 2010, Liljeroos et al., 2015, Gourlay et al., 2017). Although less accurate, pattern based approaches are over-represented in the literature due to higher complexity in development and longer computational time of the more accurate structure-based approaches (Ranganathan and Tong, 2007; Maria et al., 2017), including the sheer difference in the availability of linear versus structural data.

For a given sequence, typically all possible overlapping 9-mer (and later, 8- to 11- mer) peptide sequences are extracted. Epitope prediction for HLA class I and class II alleles are performed using benchmarked prediction models, including, for HLA class I epitopes, the artificial neural network (ANN)-trained NetMHCpan (version 4.0) (Andreatta et al., 2015, Trolle et al., 2015, Jurtz, 2017), and for HLA class II epitopes, NetMHCIIpan (Karosiene et al., 2013, Andreatta and Nielsen, 2015), and the allele-specific consensus percentile ranks of all algorithms queried by the Immune Epitope Database and Analysis Resource (IEDB) tools (combination of NN-align, SMM-align, and CombLib/Sturniolo) (Paul et al., 2015a, Paul et al., 2015b). Additionally, proteasome cleavage (Hakenberg et al., 2003, Nussbaum et al., 2001, Nielsen et al., 2005) and TAP binding predictions (Zhang et al., 2006, Bhasin et al., 2007, Bhasin and Raghava, 2004) will be performed for priority ranking. Several tools, such as NetCTL (Larsen et al., 2005), integrate these various predictions into one and predict for HLA supertypes, which are groups of HLA alleles with similar peptide binding specificity. Potentially cross-reactive self epitopes may be searched within the human proteome using BLAST/BLAT alignment tools (Altschul et al., 1990, Kent, 2002).

Case Studies: Vaccine on Infectious Diseases and Non-Infectious Diseases

Infectious Diseases

Reverse vaccinology immunoinformatics approaches are widely applied for viral vaccine design, such as for influenza virus, chikungunya virus, zika virus and others (Gupta et al., 2016, María et al., 2017), including parasites (Damfo et al., 2017) and bacteria (Mistry and Flower, 2017, Zahroh et al., 2016, Rappuoli, 2001). Khan et al. developed a bioinformatics pipeline for DENV, which proved generic as it was successfully applied to several viruses, such as WNV (Koo et al., 2009), a close relative of DENV (Khan et al., 2008), and a number of other viruses, such as HIV-1 (Hu et al., 2013), among others. It provides a novel and generalized approach to the formulation of peptide-based vaccines targeting a broad diversity of pathogens and applicable to the human population at large. This methodology is a significant contribution to the field of reverse vaccinology as it enables the systematic screening and analyses of pathogen data which would otherwise be impossible to carry out experimentally, due to too many pathogen sequences (high viral diversity) and variations in immune system among individuals (extensive polymorphism of HLA). This approach therefore significantly reduces the efforts and cost of experimentation, while providing for systematic screening and analyses of pathogen proteomes (Raman et al., 2014).

Khan et al. (2008), Koo et al. (2009) and Hu et al. (2013) analyzed a large number of dengue (DENV), West Nile virus (WNV) and clade B HIV-1, sequences, respectively, retrieved from the NCBI Entrez protein database (Table 1 ). The sequences were aligned and the overlapping nonamer amino acid positions of the viral proteome, each a possible core binding domain for human leukocyte antigen molecules and T-cell receptors, were quantitatively analyzed. The mean entropy of DENV nonamer sequences was low, with a range of 0.2–1.0 for within and 1.6–2.6 for between serotypes. This was even lower for WNV, ranging from 0.2– 0.5, with the highest for HIV-1 clade B subtype, 1.9–4.2. Entropy is a general measure of diversity, and the data provided a holistic overview viral diversity across the proteome. Accordingly, the incidence of variants to the most prevalent, index sequence at the aligned nonamer positions was the lowest for WNV (intra: ≤10%) and the highest for HIV-1 clade B (intra: ~80%–99%); the variants incidence within each DENV serotype was comparable to WNV, but between serotypes (~60%–80%) was closer to HIV-1 clade B subtype. Forty-four (44) sequences (pan-DENV sequences) identical in 80% or more of all recorded DENV sequences represented 15% of the DENV polyprotein length. The proportion (34%) was much higher for WNV and at complete conservation (100% incidence). Notably, at similar incidence level (>=80% incidence) to DENV, although pan-clade, ~35% of the intra HIV-1 clade B proteome was highly conserved. The proportion of these conserved sequences that were immune-relevant showed an inverse relationship: DENV (59% matched 9aa or more of 45 class I and II reported epitopes), WNV (50% matched 9aa or more of 57 class I and II reported epitopes), and HIV-1 clade B (37% matched 9aa or more of 73 class I and II reported epitopes). Khan et al. (2008) highlighted that conservation analysis should go beyond the species of interest, extending to all those other species that are evolutionarily related as they may act as variants to the conserved epitopes identified. This step is necessary to identify conserved epitope sequences that are pathogen specific, with none or minimal number of variant sequences within or across other pathogen species. Variant epitopes are hypothesized to cause deleterious immune responses. Many of the conserved sequences matched nine consecutive amino acids of many (flaviviruses; family of WNV and DENV) to few (deltaviruses; genus of HIV-1) other related viruses, leaving only 17, 21 and 4 pathogen specific conserved sequences for DENV, WNV and HIV-1 clade B subtype, respectively.

Table 1.

Key summary results of vaccine target discovery study on Dengue virus (DENV), West Nile virus (WNV) and clade B HIV-1.

Category	DENV	WNV	HIV-1 clade B
	(Khan et al., 2008)	(Koo et al., 2009)	(Hu et al., 2013)
Mean nonamer entropy	Intra: 0.2–1.0	Intra: 0.23–0.51	Intra: 1.9–4.2
	Inter: 1.6–2.6
General variants incidence trend	Inter: ~60%–80%	lntra: <10%	Intra: ~80%–99%
% of the proteome covered by conserved sequences	Pan-clade: 15% (>=80% incidence)	Intra: 34% (=100 % incidence)	Intra: 35% (>=80% incidence)
Immune relevance	(26/44) 59% matched 9aa or more of 45 class I and II reported epitopes	(44/88) 50% matched 9aa or more of 57 class I and II reported epitopes	(29/78) 37% matched 9aa or more of 73 class I and II reported epitopes
No. of other viral species matched by the conserved sequences	(27/44) 61% matched 64 flaviviruses	(67/88) 76% of the sequences matched 68 flaviviruses	(74/78) 95% matched 9 deltavi ruses
No. of pathogen specific conserved sequences	17	21	4

Non-Infectious Diseases

According to World Health Organization (WHO), non-infectious diseases, especially chronic diseases will lead the disability by 2020. Hence, diseases such as cancers, obesity, neurodegenerative disease addictions and others have become a recent focus of vaccine development (Barrett, 2016). Cancer is the most common non-infectious disease that leads death world-wide. Due to the advanced technology and success of in silico methods in infectious disease vaccine design, computational approaches have been applied in study of cancer vaccine design. For example, VaccImm was developed as a bioinformatics approach to simulate peptide vaccination in cancer therapy (von Eichborn et al., 2013). In addition, modeling approaches using computational biology, such as Sim Triplex and MetastaSim model are important to understand the molecular interactions at the cellular and molecular level (Pappalardo et al., 2013, Sankar et al., 2013). Adekiya et al. (2017) reports a recent example of a study that included bioinformatics analysis in cancer vaccine development.

Conclusion: Vaccine Informatics and Future Vaccines

Future vaccines will be minimalistic in approach by focusing on key parts of the pathogen, such as regions containing epitopes that cover antigenic diversity and, thus, will target immunologically similar subgroups of the human population and multiple pathogen variants. This is evident from the trend observed in evolution of vaccine strategies, which has seen a shift from whole organisms to recombinant proteins, and further towards the ultimate in minimalist vaccinology, the peptide/epitope/multi-epitope based vaccines. The minimalist approach is also expected to cover the safety concerns that are associated with the traditional vaccine approach of using whole organism (Sette and Fikes, 2003, Dertzbaugh, 1998). Vectored vaccines, suitable for ‘combination immunization’ that are produced by recombinant DNA technology and contain multivalent minimal antigens to protect against multiple infections, are considered to be the future of vaccinology (Kutzler and Weiner, 2008). The future will bring increased integration of vaccine research with advances in immunology, molecular biology, genomics, proteomics, informatics, and high-throughput instrumentation, collective defined as the emerging field of “vaccinomics”, which is hailed to be responsible for the next ‘golden age’ in vaccinology (Poland et al., 2008). Awareness of the novel technological possibilities in vaccine research is also expected to grow. Future vaccinology will be based on detailed understanding of immune function, optimal stimulation of immune responses (using adjuvants) and precise mapping and rational selection of immune targets (Brusic et al., 2005). To achieve this, vaccine development will routinely be conducted through large-scale functional studies supported by genomics, proteomics, and informatics techniques prior to clinical trials. This will provide an increased range of immune targets for vaccine design. The author expects the emergence of new generation of vaccines to be personalised to both the genetic make-up of the human population and of the disease agents. In summary, vaccinology will experience rapid progress and will eventually deliver benefits to patients from improved diagnosis, treatment and prevention of diseases.

See also

Computational Pipelines and Workflows in Bioinformatics Extraction of Immune Epitope Information Host-Pathogen Interactions Natural Language Processing Approaches in Bioinformatics

Biographies

Chong Li Chuin is a Biomedical Science undergraduate student from Faculty of Medicine and Health Sciences, UPM. As part of her undergraduate studies, she completed an internship at Perdana University Centre for Bioinformatics (PU-CBi), Malaysia. She continued to be involved with the Centre to build on her passion for bioinformatics.

Associate Professor Dr. Mohammad Asif Khan is currently the Dean of the Perdana University School of Data Sciences, the Director of Perdana University Centre for Bioinformatics (PU-CBi), Malaysia, and a Visiting Scientist at Johns Hopkins University School of Medicine (JHUSOM), USA. He obtained his PhD in Bioinformatics from the National University of Singapore (NUS), and did his postdoctoral fellowships at NUS and JHUSOM. Dr. Khan’s research interests are in the area of biological data warehousing and applications of bioinformatics to the study of immune responses, vaccines, venom toxins, drug design, and disease biomarkers.