Conformation dynamics aims at an identification of dynamically metastable subsets of the position space of molecular systems. A time-discretized molecular simulation of such a system leads to a Markov operator. A space discretization of this operator leads to a stochastic transition matrix. In the talk, a cluster algorithm is presented which identifies metastable subsets of the position space by a spectral analysis of the transition matrix. This analysis was originally valid for reversible Markov chains, but can be extended to the non-reversible case.