Genomic tools sharpen breeding value predictions for diploid banana

Banana breeders face a fundamental challenge: predicting which parent plants will produce the best offspring. A new study demonstrates that incorporating DNA marker information into statistical models significantly improves the accuracy of these predictions for yield-related traits in diploid banana. By comparing three different approaches—pedigree-based, genomic-based, and a hybrid model that combines both—the researchers found that the hybrid model delivered the most precise estimates of breeding values and heritability for traits such as bunch weight, fruit length, and fruit circumference. These findings offer a practical path toward faster, more efficient banana breeding programs.

Banana breeding is notoriously slow and labor-intensive. The crop has a long growth cycle, typically requiring two to three years of field evaluation per generation, and many plants are lost during this extended period, resulting in incomplete data. Reduced fertility in many cultivated bananas further limits the number of hybrids available for evaluation. Traditional pedigree-based methods rely on expected relationships calculated from family records, but these records are often incomplete—particularly when parental varieties are landraces with unknown ancestry. Genomic methods, which use molecular markers to estimate actual genetic sharing between individuals, offer a potential solution. Based on these challenges, there is a clear need for systematic comparison of different approaches to see which delivers the most reliable genetic predictions for banana breeding.

The research, published (DOI: 10.1093/hr/uhag139) on April 14, 2026, in Horticulture Research, was conducted by scientists from the International Institute of Tropical Agriculture (IITA) in Uganda and Tanzania, the Swedish University of Agricultural Sciences, KU Leuven in Belgium, and Gnomixx B.V. in Belgium. The team analyzed 74 diploid Mchare banana hybrids derived from a factorial cross of four female and five male parents, using 2,792 high-quality single nucleotide polymorphism (SNP) markers to construct genomic relationship matrices. They compared three models: pedigree-based best linear unbiased prediction (P-BLUP), genomic best linear unbiased prediction (G-BLUP), and a hybrid model (H-BLUP) that integrates both pedigree and genomic information.

The results revealed clear differences in how the three models partitioned genetic variance. For yield-related traits such as bunch weight, number of hands, fruit length, and fruit circumference, additive genetic variance was the dominant component across all models—but the P-BLUP model consistently overestimated these additive effects compared to the G-BLUP and H-BLUP approaches. This overestimation likely occurs because pedigree-based models fail to distinguish additive from non-additive effects, inflating the additive component. The H-BLUP model produced the smallest standard errors for variance component estimates, indicating greater precision. For prediction accuracy, the H-BLUP model outperformed the others for yield-related and cycling traits, while the P-BLUP model showed superior performance for agronomic traits such as plant height and girth. Notably, the H-BLUP method could provide predictions for traits where G-BLUP could not, because it incorporates information from both genotyped and non-genotyped individuals—a critical advantage when some plants lack DNA data due to field losses. The study also identified four top-performing hybrids with consistently high breeding values across all three models, three of which came from crosses with the female parent 'Huti-White'.

"The hybrid model gave us the best of both worlds," the authors said. "By combining pedigree records with actual DNA marker data, we could more accurately separate additive from non-additive genetic effects. This matters because additive effects are what breeders can reliably pass on to the next generation. When you overestimate additive variance, you end up with inflated expectations of genetic gain. The H-BLUP approach corrects that bias and gives breeders a much clearer picture of which parents and hybrids are truly worth advancing."

For diploid banana breeding programs, these findings translate directly into practical improvements. The H-BLUP method can be used to identify superior clones for release as new varieties or for recycling as parents in recurrent selection programs. More accurate breeding value predictions mean breeders can make better-informed decisions about which crosses to pursue, reducing the number of wasted crosses and accelerating the breeding cycle. The approach is particularly valuable in programs where pedigree records are incomplete—a common situation when working with landrace varieties. As genotyping costs continue to decline, integrating genomic information into routine breeding decisions becomes increasingly feasible, offering a pathway to more rapid genetic gains in this vital staple crop that feeds millions across East Africa and beyond.

###

References

DOI

10.1093/hr/uhag139

Original Source URL

https://doi.org/10.1093/hr/uhag139

Funding information

Gates Foundation (INV-002979, formerly OPP1213871).

About Horticulture Research

Horticulture Research is an open access journal of Nanjing Agricultural University and ranked number one in the Horticulture category of the Journal Citation Reports ™ from Clarivate, 2023. The journal is committed to publishing original research articles, reviews, perspectives, comments, correspondence articles and letters to the editor related to all major horticultural plants and disciplines, including biotechnology, breeding, cellular and molecular biology, evolution, genetics, inter-species interactions, physiology, and the origination and domestication of crops.