For the genetic model-free MDR approach, we managed to include the full SNP-dataset using an exhaustive interaction method.
For each one of the linear models, 340,768 interactions were evaluated (testing 92 against 3,704 markers), whereas for the MDR method 3,704 markers were tested in an exhaustive interaction analysis.
Finally, regardless of the statistical issues associated with exhaustive interaction testing, the computational burden is enormous - there are 1.25 × 1011 two-SNP models among 500,000 SNPs - the number typically represented on contemporary GWAS platforms.
For this purpose, we surveyed published quasi-exhaustive interaction data [ 19, 45], in which interactions between 157 mouse PDZ domains and 217 genome-encoded C-terminal peptides were tested in vitro.
The exhaustive pairwise interaction analysis showed significant differences in the interaction capacity of different Gerbera MADS domain proteins compared to other model plants.
Through exhaustive two-way interactions, IGENT_exhaust reported 39 significant interactions.
Using another approach for exhaustive search of interactions, the most recent paper by Ueki and Tamiya [ 22] also reported very few interactions in the WTCCC data.
The algorithm is computationally efficient, and thus suitable for exhaustive search for interactions along markers in a typical GWAS dataset.
However, as for PTEN, PIP3 did not interact directly with the active site, again suggesting the need for some degree of conformational change and/or more exhaustive sampling of interactions in simulations.
So, an exhaustive analysis of interaction fingerprint performance in virtual screening is presented that can aid the user to decide when and how these methods can best be employed.
However, concerns about computational efficiency remain when scaling up to exhaustive genome-wide interaction contexts.
