The interest in measuring cell-cell interactions has grown substantially in recent years. This reflects a broader shift in biology and translational research towards more biologically-relevant readouts, as well as the realization that many therapeutic mechanisms are working directly at the level of cellular interactions. A range of technologies already exists, but they differ in what they can measure, what they preserve, and what trade-offs they impose. Understanding these limitations is important, because the question is rarely whether an approach can capture interactions at all. The question is rather whether a method can do so in a way that is scalable, interpretable and compatible with therapy development workflows.

Microscopy-based or imaging-based approaches are often the most intuitive. Live-cell imaging can capture interaction dynamics directly, including contact duration, repeated engagement, and spatial organization at the immune synapse. Spatial imaging methods can preserve tissue context and reveal which cell types are positioned near each other. These approaches provide rich information, but throughput and standardization can be limiting. Imaging pipelines can be time-consuming, analysis can be complex, and comparability across experiments is often challenging. Spatial transcriptomics and related spatial omics approaches have added important capabilities. They can map gene expression within tissue context and help infer which cell types are co-localized. However, many spatial methods capture proximity rather than direct physical interaction. Therefore, they often provide indirect evidence for interaction and inference can depend strongly on resolution and cell segmentation quality. Single-cell multiomics can also provide indirect hints about cellular interactions. Transcriptional programs, surface marker expression, and activation signatures can suggest recent engagement or responsiveness to particular stimuli. Ligand-receptor inference frameworks attempt to reconstruct communication networks from expression data. These approaches can be powerful, but they remain inferential. They can suggest possible signaling relationships, but they do not directly observe physical cell-cell interactions. There are also experimental approaches that aim to preserve interacting cell pairs. Some methods capture doublets intentionally, followed by molecular profiling to assign cell identities and infer interaction partners. While promising in principle, these methods face practical challenges. Distinguishing true interacting pairs from technical artifacts can be difficult and sample handling can strongly influence observed pair frequencies. Scaling such assays while maintaining specificity remains challenging. Flow-based approaches offer a different set of trade-offs. High-throughput cytometry is compatible with many development workflows and can scale to large sample numbers. Yet standard flow cytometry typically treats doublets as artifacts and excludes them. Even when cell-cell conjugates are retained, identifying meaningful interactions reliably requires careful gating strategies and controls and biological specificity can be limited without additional design elements.

In practice, the same trade-off appears across many of these technologies. Methods that capture cell-cell interactions directly often struggle with throughput, standardization, or complexity. Methods that scale well often infer interactions indirectly or lose physical context. This does not imply that existing technologies are insufficient. Many are highly valuable within their intended scope. But it helps explain why interaction measurements have not yet become primary readouts in most immunotherapy development programs. Thus, what seems increasingly needed are approaches that combine aspects of these strengths. Technologies that preserve true interacting pairs, enable high-dimensional characterization of both partners, and scale to the sample numbers required for development decisions. Achieving this requires careful attention to sample handling, assay design and computational analysis, not only to measure interactions, but to distinguish meaningful engagement from noise and artifacts.