I've recently been looking at paper in which the authors used 'Bayesian profile analysis'. My understanding here is that they obtained MCMC output then used partitioning around medoids-type clustering (see, Section 2.2.3, and Fig.1, Pirani et al. 2015).
The authors have not made their code publicly available. However, I wondered if someone had seen similar? It would help me get my head around implementation.