Multivariate Cluster-Based Multifactor Dimensionality Reduction to Identify Genetic Interactions for Multiple Quantitative Phenotypes

Pseudocode 1

Pseudocode of multi-CMDR.

(01) perform fuzzy k-means clustering with noise cluster for phenotypes

(02) remove samples in noise cluster

(03) compute global ratio

(04) get all combinations of SNPs

(05) divide samples into N folds

(06)  for  k = 1 to N

(07)  set samples in kth folds as test dataset and the other samples as training data

(08)  for i = 1 to number of all combinations of SNPs

(09)   get all combination of genotypes

(10)    for j = 1 to number of all combination of genotypes

(11)     compute local ratio

(12)    classify each genotype combination as if , otherwise

(13)   end j

(14)   compute Hotelling’s statistics for training and test data

(15)  end i

(16)  select the best SNP combination at fold by comparing Hotelling’s statistics for training data

(17) end k

(18) compute CVC and select SNP combination with highest CVC as the best SNP combination

(19) compute p-value by permutation test for the best SNP combination