Machine learning beat traditional genetics models at predicting oyster disease resistance.

Eastern oyster farming faces a major threat from dermo disease, which is caused by a tiny parasite. In a past study, researchers tested an animal's DNA to predict disease resistance, which worked better than simply looking at physical traits. In this study, scientists gathered DNA records across three generations of oysters to improve these predictions. They tested nine different prediction systems, including three smart computer programs called machine learning models that learn patterns directly from data. They discovered that adding rare DNA markers boosted prediction accuracy for every model. Ultimately, an approach called gradient boostinggradient boostingA machine learning method that combines many simple decision rules, each one correcting the errors of the previous ones. beat every other method across all tests. It reached a top accuracy score of 0.410, easily beating the older score of 0.274.

Why it matters

Dermo disease kills farmed Eastern oysters and threatens the people who grow them. If breeders can predict which oysters carry resistance from their DNA, they can select better parents without waiting for disease to strike. This work shows machine learning, plus rare genetic variants, makes those predictions noticeably more accurate.

Questions answered