Working with the Roslin Institute, we brought down the computational time needed for a new approach to animal breeding from millions of years to a couple of hours.
Traditional horse breeders boast to know a good horse when they see one. Today, the increasing availability of genetic information is transforming the science of animal breeding. A breeder on a large industrial farm may soon have to consider thousands of genes for each of their 10,000 animals. In this context, making a breeding decision that favours desirable matches and avoids undesirable ones means accounting for millions of variables, connected by an intricate web of genetic relationships.
Researchers from Roslin Institute (best known as the birthplace of the first cloned animal, Dolly the Sheep) attempted to tackle this problem using off-the-shelf optimisation techniques. They soon found that this approach would require literally millions of years of computation time — quite a wait for lonely animals, and certainly a problem for the modern farmer who may have to make a breeding decision on a weekly basis.
Working together with the Roslin animal breeding scientists, ThinkTank Maths uncovered hidden mathematical features in the genetic information that could be exploited to bring the computational time down to a couple of hours on a normal desktop computer — paving the way for a completely new approach to animal breeding, made possible by novel mathematics.