Modelling Outcrossing of Transgenes in Maize Between Neighboring Maize Fields

Abstract

We analyze data about the flow of pollen and more specifically the outcrossing from genetically modified maze to conventional maze to determine the most important factors influencing the flow. The machine learning technique of regression tree induction is used to build models that predict the degree of outcrossing from data on the relative position of the donor and recipient fields and the winds, as well as several variables derived from these (especially considering flowering times). The resulting models show that the distance between the fields plays a dominant role, followed by the angle and the percentage of appropriate wind (blowing from the donor to the recipient field).