Multivariate models improve accuracy of genomic prediction for spring frost tolerance in Norway spruce doi-10-17044-scilifelab-28210955-0 https://doi.org/10.17044/SCILIFELAB.28210955 Swedish National Data Service Svensk nationell datatjänst Landing page Multivariate models improve accuracy of genomic prediction for spring frost tolerance in Norway spruce doi-10-17044-scilifelab-28210955-0 https://doi.org/10.17044/SCILIFELAB.28210955 Aro, Tuuli Swedish National Data Service Svensk nationell datatjänst Landing page Genomic (SNP data) and phenotype information (measurements of spring frost tolerance, bud burst and height) of the manuscript "Multivariate models improve accuracy of genomic prediction for spring frost tolerance in Norway spruce" are included in attached files. See README for more details about the files. Abstract Background Warming spring temperatures are expected to increase the risk of frost damage to emerging buds of Norway spruce by advancing their spring phenology and increasing the likelihood of damaging stochastic frost event. Current methods for assessing spring frost tolerance rely on field phenotyping after frost, which limits the ability to measure basal frost tolerance due to the sporadic nature of frost events, thereby limiting large-scale sampling. Quantitative phenotyping data are essential for genomic selection (GS) approaches that combine genomic information with phenotypic data to predict complex traits. Results In this study, we introduce a field sampling method to assess basal frost tolerance in Norway spruce using an ion leakage assay, and evaluate the effectiveness of multivariate GS models for predicting spring frost tolerance. We phenotyped 38 families across three consecutive springs at a predetermined early bud-burst stage to assess frost damage. The narrow-sense heritability of spring frost tolerance was estimated at 0.23. A strong genetic correlation between bud burst and frost damage (-0.63) improved prediction accuracy when using a multivariate model that included bud burst as an assisting trait. Additionally, we found that optimizing prediction models by increasing the number of clones in the training set further improved the accuracy of frost tolerance predictions. The observed genetic correlation between basal frost tolerance and bud burst suggests that early-flushing genotypes exhibit higher tolerance to spring frost, providing a new trait to consider in the Norway spruce breeding programs against frost damage. Conclusion The novel phenotyping method provides highly quantitative data for assessing basal frost tolerance among Norway spruce genotypes. Our study emphasizes the role of basal frost tolerance in the breeding programs of Norway spruce, particularly in the context of a warming climate, where relying solely on frost avoidance through late bud-burst timing may be insufficient. The use of multi-trait genomic prediction models for frost tolerance, incorporating assisting traits and optimized model parameters, enhances prediction accuracy and provides a promising approach for improving frost tolerance in Norway spruce. Access to data through an external actor. Åtkomst till data via extern aktör.