A stochastic simulation approach for improving response in genomic selection
The world population is increasing rapidly and is projected to hit 9.1 billion by 2050.
As the demand for food increases, agriculture production will continue to play a signiﬁcant
role. As a method to maintain and increase agriculture production, plant breeding is critical.
To improve eﬃciency in the plant breeding process, an interdisciplinary eﬀort is needed.
Operations research as a discipline focuses on decision making and eﬃcient and eﬀective
strategy design. In this thesis, operations research tools of simulation, optimization and
mathematical modeling are applied to plant breeding, speciﬁcally Genomic Selection (GS).
GS techniques allow breeders to select the best plants to make crosses by predicting, for
example, the heights of the plants using the genotypic data at an early stage of the plant
growth cycle, saving both time and cost that would otherwise be necessary to grow the
plants to maturity before their heights can be measured. A major limitation of existing GS
approaches is the trade-oﬀ between short-term genetic gains and long-term growth potential.
Some approaches focus on achieving short-term genetic gains at the cost of losing genetic
diversity for long-term gains, and others aim to maximize the long-term genetic gains but
are unable to achieve it by the breeding deadline. Our contribution is to deﬁne a new look
ahead method for assessing a selection decision, which evaluates the probability to achieve
both genetic diversity and breeding deadline. Moreover, we propose a heuristic algorithm
to ﬁnd an optimal selection decision with respect to the new method. Our new selection
method outperforms the other selection methods in the literature.