Towards a predictive understanding of influenza immunity through experimental data integration, iterative model development, and rigorous assessment of model quality

Project Summary/Abstract: Our central hypothesis is that computational models can accurately predict influenza vaccine breadth and durability based on underlying immunological factors, including an assessment of innate responses triggered by the vaccine, antigen-specific B cells and T cells, the immune exposure history of the host, and genetic factors intrinsic to the host. We propose to capture these variables and their relationship to vaccine outcomes from public datasets and newly generated experimental data. We will build computational models with complementary approaches that use these data to predict vaccine outcomes, and we will host an annual prediction contest that is open to the public to evaluate the predictive performance of these models on unseen data. Our Aims are to: 1) Capture existing influenza vaccine data and generate new experimental data that connect immunological variables with vaccine breadth and durability. We will: 1.1) Collect and standardize data from existing studies that profile the breadth and/or durability of influenza vaccine responses using any of the key immune variables we will assess (genetic factors or innate, B cell, or T cell immunity). This will result in a comprehensive training set for our computational models. 1.2) Generate new experimental datasets that measure all of these variables, along with vaccine breadth and durability, to create an independent testing set for our annual competition. 2) Generate computational models predicting the breadth and durability of vaccine responses and the cascade of immune events leading up to it. We will: 2.1) Adapt and refine computational models developed by our center investigators to predict vaccine breadth, durability, and key immune events connected to these vaccine outcomes. 2.2) Implement models published in the literature to serve as a comparison. 2.3) Combine the best- performing models to develop an integrated understanding of influenza vaccine responses. 3) Rigorously evaluate model prediction performance on unseen data in an open competition. We will 3.1) Perform an annual competition using data from Aim 1.2) that was held back from the public to evaluate model performance. 3.2) Engage the broader scientific community to participate in the predictive modeling competition and thereby maximize the diversity of independently assessed computational modeling approaches. Overall, this open, transparent, and quantitative process to build and evaluate computational models of influenza vaccination-induced immunity will test our central hypothesis and quantify how well computational models predict vaccine breadth and durability.