A high-throughput pipeline for phenotyping, object detection and quantification of leaf trichomes

Area of Science:

• Plant genetics
• Genomics
• Phenotyping technologies

Background:

• High-quality genomic data is abundant, but phenotype data acquisition remains a bottleneck for large-scale studies like genome-wide association studies (GWAS).
• Phenotype data is crucial for leveraging genomic data in association genetics, gene discovery, and validation.
• Trichomes in grasses are linked to stress tolerance, yet no trichome-related genes have been identified in this plant family.

Purpose of the Study:

• To develop a high-throughput phenotyping device and AI model for rapid collection of trichome data in wild grasses.
• To identify genomic regions associated with trichome density in Aegilops tauschii.
• To provide open-source tools for the plant science community to advance phenotyping capabilities.

Main Methods:

• Development of a portable handheld imaging device (Tricocam) for capturing leaf edge trichome images.
• Refinement and implementation of an AI-based image quantification platform for semi-automated trichome counting.
• Application of the Tricocam method and AI quantification in Aegilops tauschii, coupled with k-mer-based GWAS.

Main Results:

• Successful validation of a known trichome-controlling genomic region on chromosome arm 4DL in Aegilops tauschii.
• Discovery of a novel genomic region associated with trichome density on chromosome arm 4DS.
• Demonstration of the Tricocam and AI model's efficacy in high-throughput phenotyping for GWAS.

Conclusions:

• The developed Tricocam device and AI model significantly enhance the speed and efficiency of phenotype data acquisition for wild grasses.
• This integrated approach facilitates the identification of genetic loci controlling important traits like trichome density.
• The public release of the Tricocam design and AI model aims to empower broader applications in plant science and large-scale phenotyping projects.