Multibatch Cytometry Data Integration for Optimal Immunophenotyping

High-dimensional cytometry is a powerful technique for deciphering the immunopathological factors common to multiple individuals. However, rational comparisons of multiple batches of experiments performed on different occasions or at different sites are challenging because of batch effects. In this study, we describe the integration of multibatch cytometry datasets (iMUBAC), a flexible, scalable, and robust computational framework for unsupervised cell-type identification across multiple batches of high-dimensional cytometry datasets, even without technical replicates. After overlaying cells from multiple healthy controls across batches, iMUBAC learns batch-specific cell-type classification boundaries and identifies aberrant immunophenotypes in patient samples from multiple batches in a unified manner. We illustrate unbiased and streamlined immunophenotyping using both public and in-house mass cytometry and spectral flow cytometry datasets. The method is available as the R package iMUBAC (https://github.com/casanova-lab/iMUBAC).