Table 1 Table summarizing the assumptions made by various methodological approaches and tools for detecting epistasis
Assumption | Approach/tool |
|---|---|
Two-way interaction only | BOOST, GenEpi, EpiGWAS, MDR |
Two-, three-, or four-way interactions | BitEpi (4-way), LOBICO (4-way), MDSN (3-way) |
Any order of interaction | IRELAND, Fiuncho |
Highly polygenic/omnigenic trait | Deep neural nets |
SNP interaction partners are independent | Random forests; any tool which does not account for population structure |
Variant interactions can be described through a combination of decision trees | Random forest |
Cases can be separated from controls through a hyperplane in the (modified) variant space | Support vector machine |
Epistatic interactions are described through logic statements | LOBICO, IRELAND |