Briefly, out of 2.6 million patterns identified from the GAW15 dataset, only 10 patterns containing two or three markers were found to be statitically significant after multiple testing correction. Fully aware that false positives can still persist even after multiple testing correction, we applied a permutation test on top of the multiple testing correction to demonstrate that the false discovery rate of observing ten or more significant patterns from a similar dataset was about 2%. Two additional lines of evidence suggest that the significant multilocus associations we detected with Pattern Examiner on the data set might be true associations: 1) except for two markers (SNP0177 and SNP0672), multiple markers from the same LD region were found to be in different significant patterns; 2) all markers in the same LD region displayed consistent dominant (an allele in the pattern) or recessive (a homozygote genotype in a pattern) effect. Of course, the ultimate validation for such findings will have to come from replication studies and/or biological lab validations.

A rather skewed female/male ratio in cases vs. controls in significant patterns (in the original data the female/male ratio is 3.82/1) prompted us to investigate any potential interaction between markers in significant patterns and gender. A potentially significant interaction was observed using both chi-square test and logisticl regression.

References:

• 1. Wilcox, M.A., Li, Z., and Tapper, W. (2007). Genetic Association with Rheumatoid Arthritis – Genetic Analysis Workshop: Summary of Contributions from Group 2. Genetic Epidemiology 31(S1): S12-S21.
• 2. Li, Z., Zheng, T., Califano, A., and Floratos, A. (2007). Pattern-based Mining Strategy to Detect Multi-Locus Association and Gene-Environment Interaction. BMC Proceedings 1(Suppl 1): S16-S20.