To attempt to understand the physical principles underlying protein crystallization, an algorithm is described for simulating the crystal nucleation event computationally. The validity of the approach is supported by its ability to reproduce closely the wellknown preference of proteins for particular space group symmetries. The success of the algorithm supports a recent argument that protein crystallization is limited primarily by the entropic effects of geometric restrictions imposed during nucleation, rather than particular energetic factors. These simulations provide a new tool for attacking the problem of protein crystallization by allowing quantitative evaluation of new ideas such as the use of racemic protein mixtures.