Multifactor Dimensionality Reduction (MDR) is a statistical method designed to detect gene–gene and gene–environment interactions in genetic data. Developed by Lance W. Hahn, Marylyn D. Ritchie, and Jason H. Moore, MDR reduces high-dimensional genetic data into a single dimension, enabling the detection of interactions in relatively small sample sizes. The method uses a data reduction strategy to pool multilocus genotypes into high-risk and low-risk groups, effectively reducing the dimensionality of the genotype predictors. MDR is model-free and nonparametric, meaning it does not assume any particular genetic model or estimate parameters. It evaluates the ability of the reduced one-dimensional variable to classify and predict disease status using cross-validation and permutation testing. The method has been demonstrated to identify evidence for high-order gene–gene interactions in simulated data, sporadic breast cancer, and essential hypertension. The MDR algorithm involves partitioning data into training and testing sets, selecting genetic and environmental factors, and evaluating the ratio of affected to unaffected individuals within each multifactor cell. Cells with a ratio exceeding a threshold are labeled high-risk, while those below the threshold are labeled low-risk. The best model is selected based on misclassification error and cross-validation consistency. The MDR software can analyze interactions among 2–15 genetic and/or environmental factors, with datasets containing up to 500 variables and 4000 study subjects. The software is available in Linux and Sun versions, and is freely available to not-for-profit organizations and studies. It includes features for handling missing data and allows for the use of permutation testing to determine empirical P-values. MDR has been shown to have good power for identifying high-order interactions in simulated data and has played an important role in the identification of gene–gene interactions in real data from case-control studies. The method is currently applicable only to case-control and discordant sib-pair study designs, but future enhancements aim to expand its applicability to large, complex pedigrees. MDR is a promising approach for overcoming some of the limitations of logistic regression in detecting and characterizing gene–gene and gene–environment interactions.Multifactor Dimensionality Reduction (MDR) is a statistical method designed to detect gene–gene and gene–environment interactions in genetic data. Developed by Lance W. Hahn, Marylyn D. Ritchie, and Jason H. Moore, MDR reduces high-dimensional genetic data into a single dimension, enabling the detection of interactions in relatively small sample sizes. The method uses a data reduction strategy to pool multilocus genotypes into high-risk and low-risk groups, effectively reducing the dimensionality of the genotype predictors. MDR is model-free and nonparametric, meaning it does not assume any particular genetic model or estimate parameters. It evaluates the ability of the reduced one-dimensional variable to classify and predict disease status using cross-validation and permutation testing. The method has been demonstrated to identify evidence for high-order gene–gene interactions in simulated data, sporadic breast cancer, and essential hypertension. The MDR algorithm involves partitioning data into training and testing sets, selecting genetic and environmental factors, and evaluating the ratio of affected to unaffected individuals within each multifactor cell. Cells with a ratio exceeding a threshold are labeled high-risk, while those below the threshold are labeled low-risk. The best model is selected based on misclassification error and cross-validation consistency. The MDR software can analyze interactions among 2–15 genetic and/or environmental factors, with datasets containing up to 500 variables and 4000 study subjects. The software is available in Linux and Sun versions, and is freely available to not-for-profit organizations and studies. It includes features for handling missing data and allows for the use of permutation testing to determine empirical P-values. MDR has been shown to have good power for identifying high-order interactions in simulated data and has played an important role in the identification of gene–gene interactions in real data from case-control studies. The method is currently applicable only to case-control and discordant sib-pair study designs, but future enhancements aim to expand its applicability to large, complex pedigrees. MDR is a promising approach for overcoming some of the limitations of logistic regression in detecting and characterizing gene–gene and gene–environment interactions.