A Device for Computer Vision Analysis of Fungal Features Outperforms Quantitative Manual Microscopy by Experts in...

Area of Science:

• Plant pathology
• Genetics
• Agricultural science

Background:

• Accurate phenotyping is vital for identifying quantitative trait loci (QTL), especially those with minor effects.
• Previous methods for grapevine powdery mildew resistance phenotyping were labor-intensive, requiring significant microscopy time.

Purpose of the Study:

• To evaluate a high-throughput phenotyping device ('Blackbird') for non-destructive quantification of grapevine powdery mildew.
• To identify genetic loci associated with resistance using advanced phenotyping data.

Main Methods:

• Utilized the 'Blackbird' device combining macrophotography, robotics, and convolutional neural networks (CNNs) for image acquisition.
• Collected time-series phenotypic data from 2 to 9 days post-inoculation (dpi).
• Integrated phenotyping data with RNase H2-amplicon sequencing (rhAmpSeq) markers for genetic analysis.

Main Results:

• Blackbird achieved up to 60-fold higher throughput than manual microscopy.
• The device enabled non-destructive, time-series phenotyping of grapevine powdery mildew.
• A novel resistance locus, REN13 on chromosome 8, was detected using Blackbird-generated data.
• CNN analysis explained greater phenotypic variance (24.5% at 4 dpi for hyphae, 24.0% at 9 dpi for conidia) compared to manual microscopy (15.8% at 8 dpi).

Conclusions:

• The Blackbird device offers a rapid and efficient method for grapevine powdery mildew phenotyping.
• REN13 is a moderate-effect resistance locus in the 'Norton' variety, valuable for breeding.
• Integrating Blackbird with genetic markers facilitates QTL discovery and can enhance resistance gene stacking for durable disease management in grapevines.