Combining high-throughput deep learning phenotyping and GWAS to reveal genetic variants of fruit branch angle in upland cotton

Abstract

Cotton (Gossypium spp.), an economically and strategically significant crop in China, faces challenges such as rising cultivation costs and conflicts between grain and cotton cultivation. These challenges underscore the need for enhancing yields per unit area. In response, this study employs deep learning techniques, combined with high-throughput angle detection, to conduct genome-wide association studies (GWAS) on 355 upland cotton accessions, identify key SNPs and candidate genes for plant architecture by fruit branch angle (FBA) influences planting density, yield, and mechanized harvesting. A convolutional neural network (CNN)-based software was developed for rapid and accurate branch angle detection, showing high correlation with both AutoCAD and manual measurements. Significant phenotypic variation in FBA was observed across various cotton planting regions, with the Northwest Inland Region (NIR) exhibiting notably smaller angles. In total, 107 significant Single Nucleotide Polymorphisms (SNPs) were detected across 45 quantitative trait loci (QTL), and three potential candidate genes (Ghir_A11G034910, Ghir_D05G007790, and Ghir_D05G031350) were identified, providing insights into the genetic basis of FBA and presenting valuable genetic resources for cotton breeding programs.

Plant phenotyping relevance

CNNソフトウェアによる果枝角度の高速・高精度検出を開発し、AutoCADおよび手動測定と比較検証しているため、植物表現型取得法が中心的である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.