Polymorphisms in xenobiotic metabolizing genes are associated with altered metabolism of carcinogens in acute leukemia (AL). This study applied two data mining approaches to explore potential interactions among P53 and xenobiotic metabolizing genes in 230 AL patients [131 acute myeloid leukemia (AML) and 99 acute lymphoblastic leukemia (ALL)] and 199 controls. Individually, none of the genotypes showed significant associations with AML risk. However, in ALL the CYP1A12A TC genotype was associated with increased risk (OR = 2.02; 95% CI = 1.14-3.58; P = 0.01), whereas the GSTM1 null genotype imparted reduced risk (OR = 0.55; 95% CI = 0.31-0.96; P = 0.03). In classification and regression tree analysis, combinations of GSTM1 present, CYP1A12C AA or GG, EPHX1 exon3 TC, and EPHX1 exon4 AA or GG genotype strongly enhanced the risk of AML (OR = 5.89; 95% CI = 1.40-26.62; P = 0.01). In ALL, combinations of CYP1A12A TT, P53 GG or CC and GSTP1 AG genotypes conferred the highest risk (OR = 4.19; 95% CI = 1.45-12.25; P = 0.004). In multifactor dimensionality reduction analysis, a four locus model (GSTP1, P53, EPHX1 exon3, and CYP1A12A) was the best predictor model for ALL risk. The association between this model and ALL risk remained true even at low prior probabilities of 0.01% (false positive report probability = 0.05). Interaction entropy interpretations of the best model of ALL revealed that two-way interactions were mostly synergistic. These results suggest that high order gene-gene interactions play an important role in AL risk.