Just Accepted

This article has been peer reviewed and accepted for publication. It is in production and has not been edited, so may differ from the final published form.

The prediction of genetic values of chickpea genotypes in F4 generation by REML/BLUP and FAI-BLUP from mixed analysis models and features importance ranking by Boruta+random forest model

Sibel Ipekesen , Behiye Bicer

Abstract

Context. Selection indices are frequently used in plant breeding programs to evaluate multiple traits simultaneously. These indices allow the selection of genotypes that combine desirable characteristics for the product and show high yields. Aims. The paper aimed to predict genetic parameters of twenty-nine chickpea genotypes in the F4 generation, to determine genotypic and phenotypic correlation and genotypic and phenotypic path coefficients and to select superior chickpea genotypes by using mixed linear models. Methodology. The experiment was conducted in Diyarbakir, Southeast Anatolia of Türkiye. The experiment was arranged a randomized complete blocks design with three replications. The selection of features and superior genotypes was analyzed using some mixed analysis models, including restricted maximum likelihood-best linear unbiased prediction (REML/BLUP,) genotype-ideotype distance index (FAI-BLUP), principal component analysis (PCA), feature importance ranking (Boruta+Random Forest) and classifying features and genotypes (Ward's Clustering). Key results. These mixed analysis models effectively selected the best chickpea genotypes having advantageous genetic gains for all examined features. The selection accuracy for predicting genetic values was quite high (>96%) in the REML/BLUP model; however, it was lower (78.2%) in the PCA. In the Boruta-random forest, the superior features closely related to seed yield plant-1 were determined in chickpea genotypes. The FAI-BLUP index showed that G3, G7, G10, G23, G27, and G28 genotypes were closest to the ideotype for the features. Additionally, in Ward’s clustering analysis, these genotypes had wide genetic diversity and high similarity. Conclusions. The results of this paper indicate that the used models are effective in chickpea breeding programs to select features contributing to seed yield. Implications. This paper evaluated mixed linear models to predict genetic parameters of chickpea genotypes and recommended the best models.

CP25128  Accepted 27 August 2025

© CSIRO 2025