Gating has been the standard approach to cytometry analysis for decades, but populations are defined by pre-specified, biaxial marker combinations, so cell types outside the gating hierarchy are missed regardless of panel size. On a 43-marker panel, possible marker combinations reach into the billions, and no gating scheme can exhaustively explore them.
Unsupervised methods overcome this limit by exploring the immune landscape. Algorithms such as FlowSOM for clustering and Uniform Manifold Approximation and Projection (UMAP) for visualization identify populations directly from the data, independent of prior assumptions about which markers define which cells.
Applied to 70 PBMC samples from 29 stage II-IV melanoma patients treated with anti-PD-1 therapy, our unsupervised clustering pipeline revealed a novel population associated with response and immune-related adverse events (IRAEs) that conventional gating did not capture.
.png)
