Phenotyping: Using Machine Learning for Improved Pairwise Genotype Classification Based on Root Traits.

Area of Science:

• Agricultural Science
• Plant Biology
• Computational Biology

Background:

• Local crop cultivars are vital genetic resources, particularly for root system architecture.
• Phenotyping mature plants' root systems presents complex data analysis challenges.
• Machine learning offers promising solutions for analyzing intricate phenotyping data.

Purpose of the Study:

• To apply machine learning for unbiased root trait identification and pea cultivar classification.
• To differentiate European pea (Pisum sativum) cultivars using root system architecture.
• To assess the effectiveness of machine learning in analyzing complex phenotypic data.

Main Methods:

• Conducted a greenhouse experiment with 16 European pea cultivars in sand-filled columns.
• Collected 36 manually derived root traits for each cultivar.
• Utilized a combination of random forest and support vector machine models for analysis.

Main Results:

• Machine learning successfully identified the most distinguishing root traits.
• Up to 86% of pea cultivar pairs were differentiated using the top five important root traits (Timp5).
• Total surface area of lateral roots from tap root segments (0-5 cm depth) was a frequent key trait.

Conclusions:

• Machine learning enables unbiased trait selection and cultivar classification from complex phenotypic data.
• Root system architecture variability was largely maintained in the studied pea cultivars.
• Advanced statistical approaches are crucial for leveraging phenotyping data in crop breeding.