This document summarizes a study on clustering probabilistic graphs. It proposes extending the edit-distance-based definition of graph clustering to probabilistic graphs. Practical approximation algorithms are developed based on establishing a connection between the objective function and correlation clustering. The parameter-free objective function means the number of clusters is part of the output. Methods are developed for testing statistical significance of clusters and handling noisy clusterings. Experiments on real protein-protein interaction and large-scale social networks demonstrate the techniques can discover correct numbers of clusters and identify known relationships.