Code to accompany the paper "Optimized Permutation Testing for Information Theoretic Measures of Multi-Gene Interactions" by Kunert-Graf, Sakhanenko and Galas.
This consists of the following self-contained jupyter notebooks:
This notebook generates synthetic SNP/phenotype data and times both the naive permutation method and our method, over two different parameter ranges: first, varying the number of individuals, and then varying the number of permutations. It then plots the computation time of each approach (cf. Figure 1 of the main text).
This notebook first generates synthetic SNP/phenotype data for M individuals, and creates a total of N permuted count tables using two methods: naive permutation, and our method of transforming the count tables directly by iterative sampling of the hypergeometric distribution. It then plots the distributions of count table elements c_ij0 (cf. Figure 2 of the main text), and computes and plots the associated multi-informations (cf. Figure 3 of the main text).
This notebook generates permuted count tables using both the naive method and our method, as in both notebooks above. It then performs a 2-sample Epps Singleton test comparing the distributions of the count table entries. It does so over a range of randomly-selected allele frequencies/phenotype frequencies. It then plots the results (cf. Figure 4 of the main text).