The current  generation of genechips measure 5 million mutations per sample  contributing to a exponentially  growth of genetic information but the majority of known heritability is still unaccounted for.  The problem most likely lies in the interaction of genes with environment and genes with genes (epistasis).The traditional algorithms for calculating epistasis are exhaustive and scale quadratically with the number of mutations, linearly with the number of individuals and calculate for each mutation exactly how that mutation is correlated with any other mutation.                PAS  is a heuristic algorithm that scales linearly with the number of mutations,quadratically  with the number of individuals, for each mutation it only gives a estimate of its total participation in interactions overall not with what mutation it interacts. This makes it feasable to  combine phenotype and genotype data in a single run. The increase in data per individual is relatively inexpensive since the computational cost scales linearly with the total amount of variables the linear scaling makes it viable to use bootstrapping for hypothesis testing. The drawbacks are two fold,  no information about what flagged variables are interacting and significantly reduced detection power. Pairing the flagged variables is easy to do with traditional methods and having a weak detector that can finish in reasonable time is preferable  to having a exact tool that scales in unmanageable way.