This paper describes a scalable approach to one of the most computationally intensive problems in molecular plant breeding, that of associating quantitative traits with genetic markers. The fundamental problem is to build statistical correlations between particular loci in the genome of an individual plant and the expressed characteristics of that individual. While applied to plants in this paper, the problem generalizes to mapping genotypes to phenotypes across all biology. In this work, a formulation of a statistical approach for identifying pair wise interactions is presented. The implementation, optimization and parallelization of this approach are then presented, with scalability results.