Statistical approaches for compositional analyses


What statistical approaches are available to assess changes in cluster compositions across treatment groups? I understand this is an active area of research, and as a result, I am not finding too much on the topic. I came across one paper, scDC: single cell differential composition analysis, which uses bootstrap resampling to calculate cell-type proportions with confidence intervals. Does anyone know of other statistical approaches to this problem? If so, what is the best accepted method at the moment?


Hi, possibly related to your question, we are developing an extension to scanpy built on Stochastic Block Model approach (; there you have not only a likelihood of the clusters (and a hierarchical model of that) but a measure of cell affinity to a specific group. We are still adding/modifying the code and, more important, the documentation is not keeping the pace of development, so it is lacking some features.