The extent of genetic variation found in drug metabolism genes and its contribution to interindividual variation in response to medication remains incompletely understood. To better determine the identity and frequency of variation in 11 phase I drug metabolism genes, the exons and flanking intronic regions of the cytochrome P450 (CYP) isoenzyme genes CYP1A1, CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4 and CYP3A5 were amplified from genomic DNA and sequenced. A total of 60 kb of bi-directional sequence was generated from each of 93 human DNAs, which included Caucasian, African-American and Asian samples. There were 388 different polymorphisms identified. These included 269 non-coding, 45 synonymous and 74 non-synonymous polymorphisms. Of these, 54% were novel and included 176 non-coding, 14 synonymous and 21 non-synonymous polymorphisms. Of the novel variants observed, 85 were represented by single occurrences of the minor allele in the sample set. Much of the variation observed was from low-frequency alleles. Comparatively, these genes are variation-rich. Calculations measuring genetic diversity revealed that while the values for the individual genes are widely variable, the overall nucleotide diversity of 7.7 x 10(-4) and polymorphism parameter of 11.5 x 10(-4) are higher than those previously reported for other gene sets. Several independent measurements indicate that these genes are under selective pressure, particularly for polymorphisms corresponding to non-synonymous amino acid changes. There is relatively little difference in measurements of diversity among the ethnic groups, but there are large differences among the genes and gene subfamilies themselves. Of the three CYP subfamilies involved in phase I drug metabolism (1, 2, and 3), subfamily 2 displays the highest levels of genetic diversity.