Beyond the blood: Tissue-resident immunity shapes SARS-CoV-2 vaccine antibody responses.

In a previous study, Julia Davis-Paroda et al. presented and analyzed data from a New York organ donor tissue database to explore tissue-resident immune cells specific to SARS-CoV-2 in vaccinated and previously infected organ donors. Here we re-analyze this dataset and additional samples with various statistical and machine learning (ML) methods to investigate the relation between on the one hand SARS-CoV-2-specific T and B cell subset frequencies across tissues and in blood, and on the other hand a panel of SARS-CoV-2-specific antibody titers. We show that tissue-resident cells may be more predictive of high titers than blood cell percentages, and develop a structural equation model with imposed causal relations that makes immunological sense and is compatible with the data. One interesting statistical / data science aspect is the fact that we had to rely on multiple imputation (using the MICE R package), as the tissue sample data had many missing observations, making direct application of many of-the-shelf ML methods (RF, elastic net, CCA) infeasible.

Recommended citation: Schroeter, J. et al. (2026). "Beyond the blood: Tissue-resident immunity shapes SARS-CoV-2 vaccine antibody responses." BioRxiv.
Download Paper