Genomic selection for tar content in Nicotiana tabacum: genetic architecture analysis and model evaluation

Introduction Despite its economic importance, reducing tobacco tar content remains challenging due to its complex genetic basis. Methods Here, we evaluated 436 diverse tobacco accessions to characterize the genetic architecture of tar content and develop an optimized genomic selection strategy. Based on these findings, sixteen genomic prediction models were assessed using five-fold cross-validation. Results Genome-wide association analysis detected no major-effect loci, and regional heritability mapping revealed localized enrichment of small-effect variants, particularly on chromosome 17, indicating a predominantly polygenic architecture. rrBLUP achieved the highest prediction accuracy (0.84) with superior computational efficiency, followed closely by GBM (0.83). The robustness of rrBLUP was further confirmed in an independent panel of 36 accessions (Pearson r = 0.888). Discussion Together, our results demonstrate that tobacco tar content is governed by dispersed small-effect loci with regional aggregation and establish rrBLUP as a robust and practical model for genome-wide prediction, providing methodological guidance for low-tar tobacco breeding.