Many of the steroidogenic enzymes and cofactor proteins are bound to intracellular membranes, frustrating standard methods of structure determination. Structural models of steroidogenic P450 enzymes, however, may be predicted from the x-ray crystal structures of prokaryotic P450s. Using P450-BMP as primary structural template, models of hepatic and steroidogenic P450s have been generated using computational chemistry and graphics techniques. We have developed an analogous model of human P450c17 using an approach that relies heavily on energy minimization and molecular dynamics to yield the final structure. The final model predicts the known activities of the enzyme and explains why all reported mutations disrupt one or more activities. Although the term "computational chemistry" suggests that modeling is an operator-independent, fully automated process, modeling exercises are fraught with pitfalls, choices, and practical dilemmas which make each attempt a unique endeavor. This paper describes the procedure in detail, using P450c17 as an example, and highlights the opportunities that computational chemistry offers for the study of sex steroid biosynthesis.