Big Data and Industrial Research in Healthcare

Vaccines represent one of the most successful public health intervention to improve the quality of life and prevent life-threatening diseases. The eradication of smallpox and the substantial reduction in the incidence of poliomyelitis, hepatitis, measles, mumps, diphtheria, tetanus and meningitis have largely demonstrated that vaccination is a very cost-effective method for preventing, managing and even eradicating a disease. It has been estimated that between 2011 and 2020 approximately 20 million deaths were avoided. In the past, the traditional way of vaccine development was essentially an empirical method in which microorganisms were cultured, inactivated and injected in animals and the elicited immune response were then scrutinized with a number of immunoassays useful to identify promising vaccine target candidates. The selection of immunogenic antigens to be included in vaccines resulted after a long screening process, which was time-consuming, inherently expensive, and often paralleled by a high burden of failure. However, it is now well accepted that vaccine development process is a rather complex workflow that requires integration of information from multiple and heterogeneous areas and technologies. They include knowledge on the biology, genomics and epidemiology of the etiologic agent to be targeted by the vaccine and its infection mechanism, understanding of the cell mediated and humoral immune responses elicited during natural infection and correlated with protective immunity. Once identified, the biological role of vaccine's target molecule should be unraveled. Moreover, other important aspects that influence vaccine efficacy are related to the mode of action of typical components of vaccine formulations, like adjuvants, co-stimulatory molecules or delivery systems. Finally, considerations on reactogenicity/safety of the vaccines strongly influence the development process. In the last two decades, such high level complexity has been addressed by a growing interest in OMICs and high throughput technologies generating huge amount of data. Indeed, the recent advancements of high throughput proteomics, high resolution genomics and transcriptomics, structural biology, sophisticated bioinformatics tools combined to multi-parametric cellular immunology provide important opportunities to improve our understanding of the molecular mechanisms that underpin vaccine-mediated protection. Even more powerful are approaches integrating multiple OMICs, a process also known as systems biology, which have opened new opportunities for rationalizing vaccine target identification and for speeding up preclinical vaccines studies.

Concerning the selection of vaccine targets, accumulating evidence clearly indicate that potential candidates should fall in more than one of the following categories: (i) Secreted or membrane associated antigens. This is particularly important when antibodies are the primary mediators of vaccine-induced immunity vaccination; (ii) Abundantly expressed; for infectious diseases vaccine, such expression should be rather constant throughout the natural lifespan of the pathogen and particularly during host invasion; (iii) Conserved among epidemiologically relevant serogroups; (iv) Involved in relevant biological processes; for pathogens, toxin or virulence factors are preferred antigens. Approximately 20 years ago, genome sequencing really transformed vaccinology by allowing vaccine target selection starting directly from bacterial genome information. This strategy, termed Reverse Vaccinology, was pioneered by Rappuoli R. and collaborators who established a real vaccine discovery platform: the combination of genomics and bioinformatics is applied to identify the bacterial surface exposed/secreted proteins to be cloned, purified and tested in surrogate in vitro studies useful to predict a protective immune response. The approach was applied for the first time to the development of a vaccine against Neisseria meningitidis serogroup B (MenB), one of the major cause of bacterial sepsis and meningitis in children and young adults. Bexxero, a multicomponent broad coverage vaccine originated from this study, approved and commercialized in different countries. Later on, the complementation of Reverse Vaccinology with proteomics technologies approaches in different vaccine research programs, such as on Group B Streptococcus, Group A Streptococcus, and Chlamydia C. trachomatis, was instrumental to further refine the selection of potential vaccine candidates and rationalize downstream expensive in vivo efficacy assays. One interesting study showed that this approach can also be used for the identification of antigens that stimulate T cell responses. In this studies, an initial bioinformatic selection of the membrane-associated proteins was combined with protein array screening of human sera from individuals infected by different virulent serovars to identify the immunogenic antigens. Moreover, mass spectrometry analysis was also used for the identification of antigens expressed on the bacterial surface in different pathogenic serovars, an information which would improve the likelihood of eliciting broad coverage immune responses.

More recently, systems biology approaches integrating data from multiple OMICs and advanced bioinformatics methods have been successfully employed not only for vaccine antigen identification but also to predict the specific immune responses that correlate with protective immunity. This way, the burden of data to be managed grew up enormously. One essential aspect that needs to be addressed before starting a vaccine development program consists in the understanding of the natural infection mechanisms and the evoked immune responses able to effectively eliminate the pathogen and induce long lasting protection against re-infection. Since some pathogens are capable of manipulating the immune systems, it is important to select vaccine able to overwhelm these diverting mechanisms. Moreover, of particular significance are early gene signatures elicited days or even hours after vaccination that could exploited as novel correlates of protection or reveal mechanisms that are critical in the elicitation of the appropriate immune response, and possibly even optimize the vaccination regimen. A number of studies addressed these aspects. For instance, a systematic analysis of published transcriptional profile datasets involving 77 different host-pathogen interactions allowed to identify, shared host signatures induced in different cell types in response to different pathogens, as well as specific responses. The study also described early and late transcriptional signatures associated to antigen presenting cells during viral or bacterial infection. Other emblematic studies are those on influenza vaccines. Immunobiology events and molecular profiles underlying symptomatic influenza virus infections and signatures predictive of vaccine immunogenicity have been recently reviewed. Transcriptomics data have been used to describe specific immune signatures for live attenuated vaccines and trivalent inactivated flu vaccines. In addition, early gene signatures elicited upon vaccination with different flu vaccine formulations were identified. Similar approaches were also applied on Yellow fever as well as for pathogens causing major infectious diseases such as Plasmodium falciparum, human immunodeficiency virus (HIV), Mycobacterium tuberculosis.

Approaches exploiting data from Next Generation Sequencing (NGS), quantitative mass spectrometry, novel single cell sorting technologies offer a unique opportunity to understand the complex cellular and molecular interplay underlying the elicitation of B and T cell responses. In addition, mass cytometry (CyTOF) technologies allow to integrate multi-OMICs data with phenotypic information from different immune cell subpopulations. These holistic approaches have been used to describe the complete B cell repertoire and the T cell profiles induced in response to infection or vaccination. For vaccines conferring protective immunity by elicitation pathogen-specific antibodies, which represent the large majority on all licensed vaccines, the isolation and characterization of the antibody repertoire produced by antigen-specific B cells has acquired a central importance. This process also provides an accurate overview of the antibody maturation process and can drive effective strategies aimed at priming B cell precursors expressing germline encoded antibodies before initiation of somatic mutations. Moreover, it is instrumental to generate of functional monoclonal antibodies with therapeutic properties and, in general, to design new vaccines. For instance, the B cell specific repertoire pattern that is associated with serum antibody responses to vaccination has been shown for the tetanus toxoid vaccine. The unraveling of the B cell repertoire elicited by protective immunization, combined to single cell sorting of antigen-specific B cells and to recombinant antibody technologies provide a powerful platform to generate functionally active recombinant human antibodies. Moreover, powerful approaches also include structural proteomics technologies, such as x-ray crystallography and cryoelectron microscopy. The integration of high resolution data generated by these technologies allow to identify the protective antigen/epitope conformation eliciting functional antibodies and consequently re-instruct or optimize the vaccine design process. Relevant findings from this type of approach have been used for the optimal design of HIV and Respiratory Syncytial Virus (RSV) vaccine antigens (for review. Finally, another interesting example to scrutinize the characteristic of antibody responses elicited by vaccination consists in the combination of serum proteomics and multiple functional / biochemical assays, by which it is possible to dissect the polyclonal humoral response elicited by vaccination, as done by Chung et al. for the HIV vaccine.

A rational development of new vaccines also requires a thorough understanding of their mode of action. Indeed, unraveling the interaction network between innate and adaptive immunity could allow to develop vaccine able to selectively target desired immune responses with expected less reactogenicity. This is particularly relevant for most modern vaccines employing highly purified subunits of pathogens or recombinant antigens that necessitate the use of adjuvants or appropriate delivery systems to enhance and prolong the desired immune responses. Various systems biology approaches were used to understand the mechanism underlying the specific immune activation induced by vaccines approved for use in humans. More recently, such approaches have been used to study and compare the mode of action of different adjuvants, leading to the identification of early molecular signatures rapidly induced in the blood after vaccination that correlate and predict a protective immune response, or where associated to a better vaccine safety. A direct comparison of different adjuvants, either approved or in clinical and preclinical development stage, led to the identification of molecular signatures, pathways and networks shared by them or otherwise exclusive. In addition, studies were also dedicated to understand cytopathic effects associated to delivery systems based on viral vectors used for decades for the administration of heterogeneous antigens, by virtue of their ability to induce effectors CD8 T cells, like attenuated vaccinia virus or poxvirus, and adenovirus.

Beside contributing to the vaccine development process, computational approaches open the way to investigate the influence of additional parameters in the responsiveness to vaccines, such as environmental and lifestyle factors, pre-existing immune status, chronic infections, metabolism and geographic localization or other non-canonical factors. For instance, pre-existing immunity, sex, or age related factors were shown to affect the response to hepatitis B and the influenza vaccine . Microbiota is also an important factor influencing immune responses elicited by vaccination. Analysis of these parameters further amplifies the variety and size of data that should be managed, integrated and, rationally interpreted to provide new knowledge in the development process, a move steps toward to personalized vaccine strategies.

Overall, systems biology approaches integrating big data have revolutionized vaccinology research and have delivered new tools to inform and accelerate the research and development process. Nevertheless, there are still areas that need additional efforts. This is the case of vaccines for which protective immunity is not correlated to the elicitation of functional antibodies, but it is based on antigen specific CD8 and CD8 T cells, and different T lymphocyte subsets, such as malaria, tuberculosis and Chlamydia infections. One possible reason for this knowledge gap may be ascribed that, for ethical and pure feasibility reasons, most studies routinely analyze the immune responses in the blood, whereas functional T cells mainly reside in the tissues were they exert their protective functions. For this cases, a rational use of animal models of infection, when available, could help identify protective molecular and cellular signatures. In addition, a thorough exploitation of data from clinical trials represent another area of future improvement. High throughput technologies should be complemented to allow modeling of molecular signatures that could be associated with protective vaccination, thus enabling to establish robust correlate of protection, to predict vaccine outcome and to monitor safety. They could take into account multiple biomarkers, immunological read-outs, as well as lifestyle and environmental variables parameters possible influencing vaccination in specific population subsets. A correct integration and interpretation of data from genetics, transcriptomics, proteomics, single cell analysis, immunogenicity, toxicology and efficacy studies could tangibly accelerate vaccine development at reduced costs, and could re-inform the initial vaccine design process. To address this objective, an important challenge consists in harmonizing, processing and analyzing big data derived from heterogeneous technologies and data sources, so as to give useful interpretation. Indeed, right from the start, OMICs approaches were paralleled by the evolution of bioinformatics tools and databases to support vaccine selection. They include tools for sequence analysis, antigen topology and epitope prediction. For instance, Vaxign was a first web-based vaccine design program based on the Reverse Vaccinology strategy. Different methods and databases for storage, mining and interrogation of big data accumulated from OMICs and from literature annotation, are in continuous development to support vaccine research. Collected data include genomics, transcriptomics, proteomics, metabolomics, functional immunology, as well as information on protective antigens, DNA vaccines, and many others. Beside these research-oriented data sources, other relevant vaccine-related databases collect data from vaccine safety and reports vaccine adverse events (VAE) from many post-licensure vaccines (such as the Vaccine Adverse Event Reporting System available at www.vaers.hhs.gov), and could facilitate the association between particular adverse events and specific vaccinations. Other research databases could help overcome bottlenecks in vaccinology.