Integrating multivariate and machine learning approaches to decipher drought tolerance in okra (Abelmoschus esculentus L.)

Abstract

Drought stress severely limits crop establishment and productivity, driving early-stage screening approaches to identify resilient genotypes under changing climatic conditions. This study evaluated drought tolerance in 15 okra ( Abelmoschus esculentus L.) genotypes at the seedling stage using polyethylene glycol (PEG) induced osmotic stress. Germination, morpho-physiological, and biochemical traits were assessed to quantify genotypic responses and identify key drought-adaptive mechanisms. Significant genotypic variation was observed for most traits, indicating substantial diversity in drought response. Drought-tolerant genotypes exhibited higher germination percentage, faster germination rate, reduced mean germination time, enhanced root growth, and improved chlorophyll stability and antioxidant activity under PEG stress. Genotypes were successfully categorized using multivariate analyses based on integrated trait performance. Although different genotypes ranked highest within individual environments, the multi-trait genotype–ideotype distance index (MGIDI) consistently selected NBO-31, NBO-10, and NBO-4 under both well-watered and drought-stressed conditions, demonstrating their superior stability and broad adaptation across contrasting moisture regimes. Complementary Random Forest analysis identified root morphological, antioxidant, and photosynthetic traits as major contributors to drought tolerance. Collectively, these results elucidate the extent and trait basis of drought tolerance among okra genotypes and provide a robust early-stage screening framework for selecting drought-resilient parental material in breeding programs.