Genome-wide association study of post-harvest physiological deterioration in cassava (Manihot esculenta Crantz) using visual and AI-powered phenotyping

Abstract

Introduction: Postharvest physiological deterioration (PPD) is a rapid and severe process in cassava that causes root discoloration and spoilage soon after harvest, limiting shelf life and commercial value of the storage roots. Although environmental interventions can temporarily delay PPD, they are impractical for large-scale use. Previous studies have focused mainly on transcriptome, proteome, and candidate-gene analyses, with a few reporting robust and independently validated quantitative trait loci (QTLs). Identifying PPD-linked-QTLs is essential to breeding PPD-tolerant cassava varieties. Methods: In this study, a genome-wide association study (GWAS) was conducted to identify genomic regions associated with PPD tolerance using both human visual scoring (VS) and an artificial intelligence (AI)-powered phenotyping method. A mapping population of 298 cassava accessions was genotyped using two platforms: DArTag mid-density panel and genotyping-by-sequencing (GBS). Results and discussion: Across both genotyping platforms, 6 significant SNPs were identified using VS dataset and 6 using the AI phenotypic dataset. Consistent associations on chromosomes 1 and 12 across both phenotyping and genotyping platforms indicate potentially robust genomic regions influencing PPD response. Overall, the AI-powered phenotyping approach presents a standardized and reproducible PPD scoring procedure over the traditional visual scoring in future breeding programs for PPD. Despite differences in marker density, both DArTag and GBS markers capture comparable insights into the genetic structure of the GWAS panel. These findings provide insights into the genetic basis of PPD in cassava and offer valuable targets for marker-assisted and genomic selection toward developing cassava varieties with delayed PPD.