Integrative analysis and modeling of human immune responses and pathologies

National Institute Of Allergy And Infectious Diseases

Investigators

Linked publications, trials & patents

Abstract

We have been primarily utilizing data generated at the Trans-NIH Center for Human Immunology (CHI) to assess the immune phenotypes of healthy individuals at baseline and after perturbations, particularly with seasonal influenza vaccination. The CHI has generated multiple types of measurements of peripheral blood mononuclear cells (PBMC), including microarray data for measuring transcript abundance, multiple panels of 15-color flow cytometry for assessing cell populations (and abundance of key markers), luminex assays for measuring serum cytokine concentrations, genome-wide genotyping, and immunological endpoints such as virus-specific antibody titers and B cell Elispots. We have successfully resolved a number of data analysis and modeling challenges in the past year and have been conducting integrative modeling projects using both in-house and public data sets. These include: 1. By utilizing vaccine perturbation data, we have developed a conceptual and methodological framework to quantify baseline and response variations at the level of genes, pathways, and cell populations in a cohort of individuals. Our framework takes advantage of such natural variations to systematically infer correlates, build predictive models of response quality after immune perturbation, and infer novel functional connections among various components in the human immune system. We have applied this framework to the influenza vaccination study utilizing antibody titer response as an exemplar endpoint. We confirmed previously known post-vaccination correlates based on gene expression and plasmablast frequency from day 7 samples. More importantly, using an approach that compensates for the influence of pre-existing serology, age and gender, we derived accurate predictive models of antibody responses using pre-vaccination data alone. This finding has obvious implications for the design of future vaccine trials and for developing a deeper understanding of the molecular and cellular parameters that contribute to robust vaccine responses. The robustness and translational potential of these findings is further emphasized by our discovery that the parameters playing the greatest role in correct response prediction are those with the most stable baseline values across individuals. This raises the prospect of monitoring immune health and predicting the quality of responses in the clinic via the evaluation of these baseline blood biomarkers. The conceptual and computational analysis framework we have developed can also be applied to systems and population level exploration in a number of medically relevant circumstances, including but not limited to the effects of drug intervention or natural disease history studies in humans. We further developed novel methods to integrate cytokine, gene expression and cell population data to infer functional linkages among different components, such as those between cytokines and genes. Further, based on connections inferred among genes and cell populations, we have derived cell-subset specific gene expression signatures, and have been using this information to infer which gene subsets are involved in diverse gene expression signatures. 2. Given the presence of African American individuals in our cohort, we developed a more accurate and robust approach to perform genetic imputation to infer the missing genotypes of ethically mixed and admixed cohorts. Our approach achieves better overall imputation accuracy by customizing the imputation panels for individual haplotype blocks across the genome. 3. We have conducted genetic analysis to link SNP variants to PBMC gene expression and cell populations. These include hypothesis driven analysis that involved well known variants as well as genome-wide associations. While signals tend to be weak given the small size of our cohort, we have been developing and applying network based approaches to infer the effects of multiple genetic variants on multiple phenotypes to boost power. We are also in the process of integrating numerous genome-wide association study data sets (GWAS) of immune relevant phenotypes (including autoimmune, metabolic and infectious diseases) with our data set to decipher the molecular, cellular and immunological underpinnings of a number of immune-mediated diseases. 4. Since we routinely use and analyze publicly available data to argument data generated in-house, we have developed a web-based framework (including both user interfaces and database components) to facilitate search, retrieval, annotating, meta-analysis, and gene-expression signature generation. At the core of our tool are interfaces for creating, annotating, and sharing (among the user community) of which groups of samples can be compared to form classical gene-expression signatures or expression difference profiles the latter can be used to integrate across data sets and studies to generate virtual perturbation profiles across all genes. Since annotating such comparison groups is often one of the most time-consuming steps of reusing existing data, our framework provides functions for users to share their own annotations and search for others annotations. Our tool was designed for experimental biologist to take full advantage of reusing and sharing large-scale data for obtaining biological insights.