Prostate cancer will be diagnosed in about 179,300 men in the US in 1999 alone. Some 37,000 individuals die of this disease annually. Prostate cancer is characterized by a substantial racial/ethnic variation in risk: highest in African-American men, lowest in Asian men and intermediate in Caucasian and Latino men. We set out to investigate as our central hypothesis that genetic variants of genes involved in androgen metabolism by themselves and in combination significantly contribute to prostate cancer progression and its racial/ethnic variation. Specifically, we examined the hypothesis that DNA sequence (allelic) variations in the type II (or prostatic) steroid 5alpha-reductase (SRD5A2) gene contribute substantially to the risk and progression of prostate cancer particularly across racial/ethnic lines. The "candidate gene", SRD5A2, was chosen because the reaction product [i.e. dihydrotestosterone (DHT)] of the enzyme encoded by this gene modulates directly cell division in the prostate. DHT binds to the androgen receptor (AR) and the DHT-AR complex leads to the transactivation of a variety of genes which ultimately modulates cell division in the prostate. Epidemiologic evidence suggests that variation in DHT levels play an important role in risk of prostate cancer. Thus, steroid 5alpha-reductase activity encoded by SRD5A2 variant alleles may be important in regulating intraprostatic DHT steady state levels by controlling its biosynthesis. A second candidate gene, the type II 3beta-hydroxysteroid dehydrogenase (HSD3B2) gene, encodes the enzyme that initiates the metabolic inactivation of testosterone (T) to DHT. We have identified allelic variants in this gene as well. Here I review our strategy for identifying candidate genes for prostate cancer, a multifactorial disease. I summarize the significant findings, particularly of allelic variants in the HSD3B2 and SRD5A2 genes and discuss how they by themselves, in combination and through interactions with the environment may play a role in prostate cancer predisposition and its progression. Our approach, a multidisciplinary genomic genetic (GEN GEN) attack on the problem, may be useful in the analysis of other complex phenotypes as well.