Hybrid genomic model boosts banana breeding predictions

Banana breeders have long faced a stubborn problem: figuring out which parent plants will produce the best offspring. The crop's biology works against them. A single generation takes two to three years of field work to evaluate, many plants die before the cycle completes, and reduced fertility in most cultivated varieties means fewer hybrids to test. Pedigree records—the traditional tool for predicting which crosses might succeed—are often incomplete or missing entirely, especially when working with landrace varieties whose ancestry is simply unknown. A new study published in April 2026 in Horticulture Research suggests that combining DNA marker data with those incomplete family records produces significantly more accurate predictions than either approach alone.

Researchers from the International Institute of Tropical Agriculture in Uganda and Tanzania, the Swedish University of Agricultural Sciences, KU Leuven in Belgium, and Gnomixx B.V. tested three different statistical models on 74 diploid Mchare banana hybrids. They used 2,792 high-quality genetic markers to build what's called a genomic relationship matrix—essentially a map of actual genetic similarity between plants, derived from their DNA rather than guesswork about family connections. The three models they compared were straightforward in concept: one based purely on pedigree records, one based purely on genomic data, and a hybrid that used both sources of information together.

The results showed clear patterns. When predicting traits tied to yield—bunch weight, number of hands, fruit length, fruit circumference—the hybrid model consistently outperformed the other two. It produced smaller standard errors, meaning its estimates were more precise. The pedigree-only model, by contrast, systematically overestimated how much of the variation in these traits came from additive genetic effects—the kind of effects that breeders can reliably pass to the next generation. This overestimation happens because pedigree models can't distinguish between additive effects and non-additive ones, so they lump everything into the additive category and inflate it. The hybrid model corrected that bias.

There was another practical advantage. The hybrid model could generate predictions for plants that had never been genotyped—a real benefit in breeding programs where field losses mean some individuals never make it to the lab. The genomic-only model couldn't do this; it needed DNA data on every plant. For a few agronomic traits like plant height and girth, the pedigree-only model actually performed better, suggesting that different traits may respond differently to different prediction methods. But for the traits breeders care most about—yield and cycling time—the hybrid approach was superior.

The study identified four top-performing hybrids with consistently strong breeding values across all three models. Three of them came from crosses using the female parent 'Huti-White,' a signal that this variety might be worth prioritizing in future breeding work. The authors noted that the hybrid model gave them "the best of both worlds." By combining family records with actual DNA data, they could separate additive from non-additive effects more accurately, which matters because overestimating additive variance leads breeders to have inflated expectations about genetic gain. A clearer picture of which parents and hybrids are truly worth advancing means fewer wasted crosses and a faster breeding cycle.

For programs working with landraces or other varieties where pedigree information is sparse or unreliable, this approach offers a direct path forward. As the cost of genotyping continues to fall, integrating genomic information into routine breeding decisions becomes increasingly practical. The hybrid model can identify superior clones ready for release as new varieties or for recycling as parents in recurrent selection programs. In a crop that feeds millions across East Africa and beyond, faster genetic gains translate directly into improved food security. The research, funded by the Gates Foundation, suggests that the bottleneck in banana breeding may finally have a systematic solution.