Novel image descriptors, along with an accurate and rapid phenotyping method, were introduced to improve pod constriction evaluation in peanut. A major consensus QTL region spanning a 728-kb interval was newly identified for pod constriction metrics. A candidate gene exhibiting a frameshift variation was identified within the consensus region and its diagnostic markers were developed. Pod constriction (PC) is a key morphological trait determining both commercial values and yield of in-shell peanuts. Conventional phenotyping metrics (visual scores and pod waist length derived descriptors) suffer from low precision or limited applicability, especially for atypical pod shapes, which have constrained discovery of underlying genes. To address these limitations, this study introduced two novel image descriptors: front and back constriction depth indices (FrontDI and BackDI). These indices enable accurate and robust evaluation of PC across diverse pod morphologies. Additionally, a Python script employing the deep learning technology was developed to efficiently and precisely extract these metrics. By applying both novel and conventional phenotyping methods to a recombinant inbred line population (Luhua 11×06B16), this study identified four quantitative trait loci (QTLs) for FrontDI, four for BackDI, three for visual score, and two for a pod waist length-based descriptor across three environments. A major and co-localized QTL region was consistently detected on chromosome 2. Meta-analysis further refined this region to a 728-kb consensus interval. Within this interval, an InDel was identified in the coding region of Arahy. X14VTN between the two parental lines, resulting in a frameshift mutation and a predicted alteration in protein structure. Diagnostic markers were developed for this candidate gene, confirming the genetic effect on PC variation. The novel image descriptors and genetic loci presented here improve our understanding of the genetic basis of PC in peanut and offer practical tools for molecular breeding aimed at trait improvement.
Zhang et al. (Sun,) studied this question.