Abstract
Sugarcane mosaic virus (SCMV) is the main etiological agent of sugarcane mosaic disease, which affects sugarcane and other grass crops. Despite the extensive characterization of quantitative trait loci controlling resistance to SCMV in maize, the genetic basis of this trait in sugarcane is largely unexplored. Here, a genome-wide association study was performed and machine learning coupled with feature selection was used for genomic prediction of resistance to SCMV in a diverse sugarcane panel. Nine single-nucleotide polymorphisms (SNPs) explained up to 29.9% of the observed phenotypic variance and a 73-SNP set predicted resistance with high accuracy, precision, recall, and F1 scores (the harmonic mean of precision and recall). Both marker sets were validated in additional sugarcane genotypes, in which the SNPs explained up to 23.6% of the phenotypic variation and predicted resistance with a maximum accuracy of 69.1%. Synteny analyses suggested that the gene responsible for the majority of SCMV resistance in maize is absent in sugarcane, explaining why this major resistance source has not been identified in this crop. Finally, using sugarcane RNA-Seq data, markers associated with resistance to SCMV were annotated, and a gene coexpression network was constructed to identify the predicted biological processes involved in resistance. This network allowed the identification of candidate resistance genes and confirmed the involvement of stress responses, photosynthesis, and the regulation of transcription and translation in resistance to SCMV. These results provide a practical marker-assisted breeding approach for sugarcane and identify target genes for future studies of SCMV resistance.
