In plant breeding, genomic selection is a powerful tool allowing for early and accurate selection of complex traits, reducing the time and resources required for breeding cycles. However, in the quest for alternatives to genotyping, it has been proposed recently a low-cost and efficient alternative, called phenomic selection, in which genotyping is replaced by endophenotypes such as near-infrared spectra (NIRS).At INRAE Bordeaux, the PrADAM team working on Prunus aims at predict complex traits related to phenology in sweet cherry to facilitate the early selection of interesting hybrids in breeding programs. Therefore, a comparison of the predictive ability between genomic and phenomic predictions would represent a step forward in sweet cherry breeding. Using sweet cherry germplasm collection of INRAE Prunus-Juglans Biological Resources Center, we investigated both approaches: (i) genomic prediction based on 115 accessions phenotyped for beginning, full and end flowering dates, and genotyped using the 6K SNP Illumina Infinium® array, and (ii) phenomic prediction based on 98 accessions phenotyped for the same traits using Bruker® NIRS equipment.Preliminary results gave satisfying predictive ability for both approaches. Indeed, using rrBLUP model, 1,294 SNPs and BLUPs over 7 years of phenotyping, we obtained genomic predictive abilities (PA) of 0.61 to 0.66 depending on the trait, using a training set size of 80%. Using NIRS data as phenomic tool, we observed an effect of the preprocessing method on PA such as Savitzky–Golay (SG) derivatives, standard normal variate (SNV) or multiplicative scatter correction (MSC). Using rrBLUP and one single year of phenotyping, the best PA of 0.52 was obtained for full flowering date with the combination of SNV and first SG derivative, using a training set size of 80% as well. This study highlights the promising opportunity in the use of such new omics-based breeding approaches for complex traits in sweet cherry.< Leer menos