Molecular dosimetry for polycyclic aromatic hydrocarbon-DNA adducts, genetic predisposition to cancer, and their interrelationships are under study in numerous laboratories. This report describes a modified 32P-postlabeling assay for the detection of polycyclic aromatic hydrocarbon-DNA adducts that uses immunoaffinity chromatography to enhance chemical specificity and quantitative reliability. The assay incorporates internal standards to determine direct molar ratios of adducts to unmodified nucleotides and to assess T4 polynucleotide kinase labeling efficiency. High performance liquid chromatography is used to assure adequacy of DNA enzymatic digestion. The assay was validated using radiolabeled benzo(a)pyrene-diol-epoxide modified DNA (r = 0.76, P < 0.05) thereby assessing all variables from enzymatic digestion to detection. Thirty-eight human lung samples were examined and adducts were detected in seven. A subset of samples also was examined for benzo(a)pyrene-diol-epoxide-DNA adducts by immunoaffinity chromatography, high performance liquid chromatography, and synchronous fluorescence spectroscopy. A high correlation between the two assays was found (P = 0.006). The lung samples were then analyzed by the polymerase chain reaction for the presence of mutations in the cytochrome P-450 (CYP) 1A1 and glutathione S-transferase mu (GST mu) genes. A positive association was identified for adduct levels and GST mu null genotypes (P = 0.038). No correlation was found between polycyclic aromatic hydrocarbon-adduct levels and CYP1A1 exon 7 mutations. Age, race, and serum cotinine were not related to adduct levels. Multivariate analysis indicated that only the GST mu genotype was associated with polycyclic aromatic hydrocarbon-DNA adduct levels. This work demonstrates that the 32P-postlabeling assay can be modified for chemically specific adduct detection and that it can be used in the assessment of potentially important genetic factors for cancer risk. The absence of a functional GST mu gene in humans is likely one such factor.