Journal of Agronomy and Field Innovations

Chickpea (Cicer arietinum L.) is a critical grain legume providing sustainable dietary protein to millions across arid and semi-arid regions. However, terminal drought and accelerating climate variability present severe threats to global yields. Developing varieties characterized by both early maturity—to escape late-season drought—and long-term yield stability across erratic environments is a primary agricultural directive.
Traditional phenotypic selection handles complex polygenic traits inefficiently due to substantial Genotype-by-Environment (G×E) interactions. This study implements Genomic Selection (GS) across a diverse training population of 360 chickpea genotypes evaluated over three years across four distinct agro-ecological locations. Genotyping was performed via high-density Single Nucleotide Polymorphism (SNP) arrays, yielding 14,850 high-quality markers after stringent filtration. We evaluated five advanced predictive models: Ridge Regression Best Linear Unbiased Prediction (rrBLUP), Genomic Best Linear Unbiased Prediction (GBLUP), Bayesian LASSO (BL), Bayesian B (BB), and Random Forest (RF).
Prediction accuracies for early maturity (days to 50% flowering and days to maturity) reached high levels across all parametric models. Yield stability metrics, derived via Finlay-Wilkinson and AMMI stability parameters, showed moderate to high predictability. Incorporating historical multi-environment structural components into GBLUP architectures boosted cross-validated prediction accuracies by 18% compared to single-environment models. These findings demonstrate that genomic selection can compress conventional breeding cycles by up to 50%, enabling rapid selection of resilient, short-duration chickpea cohorts.