Email: e.dinglasan@uq.edu.au

As a standard practice, parents for crossing are selected based on their performance per se or their breeding value. A significant challenge for elite wheat breeding germplasm is to increase rates of genetic gain without exhausting the genetic diversity. This is particularly challenging when a combination of quantitative traits (yield, grain quality, heat-tolerance) are considered simultaneously.

A new optimal haplotype selection strategy – called ‘FastStack’ that uses evolutionary computing, a type of AI optimisation technique, was developed at UQ, in partnership with LongReach, to first identify and then stack desirable chromosome segments in the shortest possible time. In our pilot study to improve wheat yield, we used genetic algorithm in a large commercial data set comprising more than 40,000 genotyped breeding lines which have been tested over multiple years (2010–2018) and locations across Australia.

The objective function of the algorithm predicts optimal crosses that most efficiently stack complementary haplotypes for yield; and generated offspring from the digital twin is advanced through speed breeding which allows a rapid turnaround of several generations and breeding cycles per year.

For the first time, validation trials (2022–2023) showed top performing breeding lines have considerably better yield than existing varieties. Selected lines are also now being further advanced (2024) and/or used as new parents to introgress stacks into LongReach elite germplasm.

With the success of our pilot study, LongReach is now in a position to integrate the AI-guided optimal haplotype stacking strategy within their breeding program. To expedite this development further, we are also currently developing a multi-trait approach on how to stack desirable haplotypes for grain quality and heat-tolerance into high yielding backgrounds.