Fig. 2

Pipeline of BHIT high-order interaction detection. Firstly the pipeline preprocesses raw genotype data by missing data imputation and MAF filtering, and then checks raw continuous phenotype data to confirm whether it follows the normal distribution by the Kolmogorov–Smirnov test. Both genotype and phenotype data should be combined together and converted to the BHIT file format. BHIT is applied upon these data via different strategies and validates its results in the end