A novel strategy for plant breeding based on simulations of gene network models

This paper describes a new computational method to make predictions on the outcome of pair-wise crosses of plant lines limiting expensive laboratory breeding experiments to carry out crosses of the most promising pairs of lines. Compared to the well-known marker assisted breeding, the proposed approach approximates plant gene regulatory networks to estimate outcomes of all possible crossovers, thereby taking into account epistatic relationships between alleles. The proposed method is tested and compare with various breeding approaches on artificial NK landscape models and an extensive synthetic model of Arabidopsis plant's flowering time system. The results show our method outperforms other breeding strategies.