We note that the bridge integration is particularly well-suited for experimental designs where multi-omic technologies can be applied to a subset, rather than all, experimental samples. This is a common occurrence, particularly because multi-omic technologies often are associated with increased cost, lower throughput, and reduced data quality for each individual measurement type. In particular, we note that combinatorial indexing approaches can be readily applied using commercial instrumentation to profile a single modality in hundreds of thousands of cells74,75, but the same is not true for multi-omic technologies. We propose that the collection of large single-modality datasets, harmonized via a smaller but representative multi-omic bridge, may represent an efficient and robust strategy to explore cross-modality relationships across millions of cells. Our identification of cell cycle ‘priming’ in hematopoietic stem cells represents an example of cross-modality insights that can be derived via bridge integration. We note that since cells are integrated into a shared latent space, we can characterize these relationships based on the original measurements, and do not need to perform cross-modality imputation.
