Harnessing Matrix Completion to Unify and Extend Viral Serology Studies

Abstract

The development of new vaccine strategies, as well as our understanding of key processes that shape viral evolution and host antibody repertoires, rely on measuring multiple antibody responses against large panels of viruses. Given the enormous diversity of circulating virus strains and antibody responses, exhaustive testing of all antibody-virus interactions is infeasible. Even within individual studies with limited panels, exhaustive testing is not always done, and there is no common framework for combining information across studies with partially overlapping panels, especially when the assay type or host species differ. Results from prior studies have demonstrated that virus-antibody interactions can be characterized in a vastly simpler and lower-dimensional space, suggesting that relatively few measurements could accurately predict unmeasured antibody-virus interactions. Here, we apply matrix completion to several of the largest-scale studies for both influenza and HIV-1. We explore how prediction accuracy evolves as the number of available measurements changes and approximate the number of additional measurements necessary in several highly incomplete datasets (suggesting [~]250,000 measurements could be saved). In addition, we show how the method can combine disparate datasets, even when the number of available measurements is below the theoretical limit that guarantees successful prediction. This approach can be readily generalized to other viruses or more broadly to other low-dimensional biological datasets. SignificanceOne of the central problems in immunology is to characterize how our vast array of antibodies inhibits the diverse pathogens we encounter in our lives. In this work, we apply a well-studied mathematical technique called matrix completion that leverages patterns in partially-observed antibody-virus inhibition data to infer unmeasured interactions. We predict the results of tens of thousands of missing experiments in influenza and HIV-1 studies and quantify the expected error of our estimates. By harnessing matrix completion, future experiments could be designed that only collect a fraction of measurements, saving time and resources while maximizing the information gained.