I will present a clustering algorithm for graphs called the Markov Cluster
Algorithm (MCL). The algorithm is the basic interface to an algebraic
process defined on stochastic matrices, namely, the MCL process. Let G be
a weighted and directed graph. The MCL algorithm simulates flow in G by
flow expansion and contraction. The heuristic underlying this method is
that the flow between dense regions which are sparsely connected is
expected to evaporate.

The algorithm is part of a PhD thesis by Stijn van Dongen.
http://micans.org/mcl/lit/

The algorithm can be used for large-scale detection of protein families.
If interested, refer to:

An efficient algorithm for large-scale detection of protein families A. J.
Enright, S. Van Dongen, and C. A. Ouzounis. Nucleic Acids Res. 2002 April
1; 30(7): 1575?1584.