Identifying graft incompatible rootstocks for sweet cherry through machine learning algorithms.

Area of Science:

• Horticulture
• Plant Breeding
• Genetics

Background:

• Graft incompatibility is a significant challenge in developing dwarf and semi-dwarf sweet cherry (Prunus avium L.) rootstocks.
• Optimizing rootstocks is crucial for enhancing yield, fruit quality, precocity, and labor efficiency in cherry cultivation.

Purpose of the Study:

• To evaluate the graft incompatibility of eight native cherry and mahaleb genotypes from Northern Anatolia.
• To identify compatible rootstock candidates for sweet cherry using a multidisciplinary approach.

Main Methods:

• Grafting eight selected genotypes and standard rootstocks (Gisela 6, SL 64) with '0900 Ziraat' and 'Lambert' cultivars.
• Assessing graft incompatibility through morphological (bud growth, shoot length) and anatomical parameters, alongside advanced data-driven analyses (PCA, Random Forest, SHAP, Bayesian ranking).

Main Results:

• Key parameters like graft bud growth rate (40.26-86.21%), shoot length (41.01-91.28 cm), and rootstock/scion diameter ratio (0.41-0.92) were measured 12 months post-grafting.
• The integrated analysis successfully identified five genotypes exhibiting high compatibility potential with sweet cherry cultivars.
• Machine learning models effectively ranked genotype performance and identified critical traits influencing compatibility.

Conclusions:

• Combining traditional phenotypic evaluations with machine learning provides a robust framework for assessing graft incompatibility in sweet cherry.
• The identified compatible genotypes offer valuable resources for future sweet cherry breeding programs and rootstock selection strategies.