“Omitted” Cell Detection in BulkTrajBlend: Here, a neighborhood graph constructed via UMAP based on the generated single-cell data identifies nodes corresponding to individual cells and delineates distinct communities by cell type. The annotated graph is the input for a Graph Neural Network (GNN) that detects overlapping communities and identifies mixed cell types, which are then reintegrated into the original single-cell profile. Overlap-CellType: a one-hot matrix of cell types for the overlapping Celltypes.
