Seurat’s mixing metric (mm)

Metric using the median position of the kth cell from each batch within its k-nearest neighbours (knn) as a score. In general the lower the metrics score the better mixed is the neighbourhood, but it’s absolute range effectivly depends on the number of batches, k and the size of knn, e.g. even with constant k (default = 5) and constant size of knn (default = 300) the metric’s minimum is 10 for a dataset with 3 batches and 15 for a dataset with 5 batches. This makes it less suitable for an across dataset comparison, but still valid for within dataset compartisons.

The mixing metric has been suggested by Stuart et al., 2018 and is available in the Seurat R package as well as in the Bioconductor CellMixS R package.