Periodic Reporting for period 2 - FluPRINT (Tracing the inFLUenza vaccine imPRINT on immune system to identify cellular signature of protection)
Período documentado: 2020-06-01 hasta 2021-05-31
The mechanisms of how protective immunity against influenza is accomplished and which components of the immune system are necessary to mount an effective response to influenza are currently unclear. Uncovering these mechanisms would help to improve current vaccines. The main goal of this project was to characterize influenza vaccine-induced immune responses with the aim of defining cellular and molecular correlates of protection. This project covered an issue that has been poorly studied in humans and that is the role of influenza-specific T cells after vaccination. Correlating the cellular signature after vaccination with the vaccine efficacy is a novel approach to the current problem about the usage of influenza vaccines.
Recent advances in computational biology make it possible to extract knowledge and identify patterns in an unbiased manner from large clinical datasets and to integrate different data types collected across studies. In this work, we have developed a novel computational approach that automates data analysis, Sequential Iterative Modeling “OverNight” (SIMON). Our approach automatically builds state-of-the-art machine learning models testing more than 180 algorithms to find the ones which fit any given data distribution. Such an iterative process maximizes predictive accuracy of the generated models and is especially suited for clinical data collected across multiple cohorts. SIMON was applied to data from five clinical studies across eight influenza seasons and with over 3,000 parameters considered. SIMON identified several immune cell subsets, including CD4+ T helper cells, regulatory T cells, and cytotoxic CD8+ T cells that correlated with an effective antibody response to influenza vaccination. While T helper cells are in general thought to be the principal source of T cell “help” for antibody production and the generation of high-affinity memory B cells, the role of regulatory and cytotoxic T cells in the induction of antibodies is unknown.
Experimental data confirmed that these cell subsets were elevated in vaccine responders.
Altogether, our findings reveal the unexpected role of cytotoxic and regulatory T cells in the generation of protective antibody responses after influenza vaccination, as well as provide a novel tool for the integration of multi-omics data. The ‘systems immunology’ approach established in the FluPRINT research project can accelerate the understanding of how immune responses to influenza are generated and can help to improve future flu vaccine formulations, while SIMON as an open-source knowledge discovery software it can help researchers to identify biomarkers important for other vaccines, therapies and diseases, such as it was applied to understanding correlates of protection in COVID-19 patients and durability of protective immunity after SARS-CoV-2 infection.
Recent advances in computational biology make it possible to extract knowledge and identify patterns in an unbiased manner from large clinical datasets and to integrate different data types collected across studies. In this work, we have developed a novel computational approach that automates data analysis, Sequential Iterative Modeling “OverNight” (SIMON). Our approach automatically builds state-of-the-art machine learning models testing more than 180 algorithms to find the ones which fit any given data distribution. Such an iterative process maximizes predictive accuracy of the generated models and is especially suited for clinical data collected across multiple cohorts. SIMON was applied to data from five clinical studies across eight influenza seasons and with over 3,000 parameters considered. SIMON identified several immune cell subsets, including CD4+ T helper cells, regulatory T cells, and cytotoxic CD8+ T cells that correlated with an effective antibody response to influenza vaccination. While T helper cells are in general thought to be the principal source of T cell “help” for antibody production and the generation of high-affinity memory B cells, the role of regulatory and cytotoxic T cells in the induction of antibodies is unknown.
Experimental data confirmed that these cell subsets were elevated in vaccine responders.
Altogether, our findings reveal the unexpected role of cytotoxic and regulatory T cells in the generation of protective antibody responses after influenza vaccination, as well as provide a novel tool for the integration of multi-omics data. The ‘systems immunology’ approach established in the FluPRINT research project can accelerate the understanding of how immune responses to influenza are generated and can help to improve future flu vaccine formulations, while SIMON as an open-source knowledge discovery software it can help researchers to identify biomarkers important for other vaccines, therapies and diseases, such as it was applied to understanding correlates of protection in COVID-19 patients and durability of protective immunity after SARS-CoV-2 infection.
To identify why some individuals are protected against flu following vaccination and why some are not, we developed automated machine learning approach for the integration of different data types and clinical studies. This was accomplished by the generation of the FluPRINT database, (Tomic et al, Sci Data, 2019) which consists of 13 different assays performed on blood and serum samples taken from 740 individuals undergoing influenza vaccination. The dataset contains information on more than 3,000 parameters measured using mass/flow cytometry, phosphorylation-specific cytometry, multiplex cytokine assays, clinical lab tests, serological profiling, and virological tests. Next, we used this dataset to uncover new markers and mechanisms that are important for influenza vaccine immunogenicity. To accomplish that we have developed SIMON, an automated machine learning tool that is particularly suitable for clinical data collected across multiple cohorts containing inconsistent features with many missing values. We demonstrated that such a process maximizes predictive accuracy and other performance measurements of the generated models. We have applied SIMON to the generated dataset and identified several previously unknown immune cell subsets, that correlated with a successful influenza vaccination. Overall, SIMON facilitates pattern recognition and knowledge extraction from high-dimensional clinical data collected across multiple cohorts. To read more about SIMON and how we applied it to flu vaccination studies, please check out our publications (Tomic et al, JI, 2019, https://doi.org/10.4049/jimmunol.1900033 and Tomic et al, Patterns, 2021) or the project's website at https://fluprint.com/.
The final results of the FluPRINT project contributed to our knowledge about the effectiveness of influenza vaccines by providing an integrated picture of the mechanistic details associated with suboptimal vaccine effectiveness which could drive the development of the next generation of influenza vaccines.
The integration of data using machine learning and our systems immunology approach is a rapidly growing area that has strategic importance in the future offering many innovative applications in biomedicine. Marie Curie fellow was able to develop unique skill sets in multi-omics technologies, including mass cytometry, sequencing and proteomics, computational biology, and machine learning. Also, fellow obtained extensive training in systems immunology with Dr. Mark Davis, who pioneered human immunology and multi-omics analysis. Marie Curie fellowship supported fellow’s work and contributions to the emerging field of systems immunology with several manuscripts and reviews, invited presentations at conferences and different institutes, providing lectures and fellow received several awards for contributions to the development of new multi-omics integration methodologies and algorithms. The impact of fellow’s research under the Marie Curie fellowship gained more attention, one of the publications was rated by the American Association of Immunologists as the most read manuscript for 2019/2020. This work was also highlighted by the University of Oxford (https://www.ox.ac.uk/research/research-impact/simon-says) Fluidigm (https://www.fluidigm.com/articles/machine-learning-and-infectious-disease) and by online immunology research website (https://immunobites.com/2019/08/05/the-game-of-simon-says-predicting-flu-vaccine-response-using-machine-learning/)%20). Fellow was also selected as a Marie Curie fellow of the week.
During Marie Curie fellowship, in collaboration with Dr. Davis at Stanford University and Dr. Pollard at the University of Oxford, fellow strengthened scientific network of scientists which resulted in several successful national and international projects, including the Oxford Immunology Consortium for COVID-19 research.
Overall, with the outstanding scientific, leadership, and management output within the FluPRINT project, fellow received invaluable experience essential for the establishment of the independent research group.
The integration of data using machine learning and our systems immunology approach is a rapidly growing area that has strategic importance in the future offering many innovative applications in biomedicine. Marie Curie fellow was able to develop unique skill sets in multi-omics technologies, including mass cytometry, sequencing and proteomics, computational biology, and machine learning. Also, fellow obtained extensive training in systems immunology with Dr. Mark Davis, who pioneered human immunology and multi-omics analysis. Marie Curie fellowship supported fellow’s work and contributions to the emerging field of systems immunology with several manuscripts and reviews, invited presentations at conferences and different institutes, providing lectures and fellow received several awards for contributions to the development of new multi-omics integration methodologies and algorithms. The impact of fellow’s research under the Marie Curie fellowship gained more attention, one of the publications was rated by the American Association of Immunologists as the most read manuscript for 2019/2020. This work was also highlighted by the University of Oxford (https://www.ox.ac.uk/research/research-impact/simon-says) Fluidigm (https://www.fluidigm.com/articles/machine-learning-and-infectious-disease) and by online immunology research website (https://immunobites.com/2019/08/05/the-game-of-simon-says-predicting-flu-vaccine-response-using-machine-learning/)%20). Fellow was also selected as a Marie Curie fellow of the week.
During Marie Curie fellowship, in collaboration with Dr. Davis at Stanford University and Dr. Pollard at the University of Oxford, fellow strengthened scientific network of scientists which resulted in several successful national and international projects, including the Oxford Immunology Consortium for COVID-19 research.
Overall, with the outstanding scientific, leadership, and management output within the FluPRINT project, fellow received invaluable experience essential for the establishment of the independent research group.