Identification and Classification of Salt-Tolerant Rice Germplasm Using Regression-Based Phenotyping and Multivariate Cluster Analysis

Salinity is a major constraint on rice production, and reliable screening methods are needed to identify germplasm with favorable responses across different stress levels. This study evaluated seedling-stage salinity responses in 210 rice accessions exposed to 0, 0.25, 0.50, and 0.75% NaCl under hydroponic conditions. Total length, shoot length, root length, dry weight, and visual salt-injury score were measured, and accession-specific regression coefficients were calculated to quantify concentration-dependent responses. Increasing NaCl concentrations progressively reduced all growth traits and increased visible injury. At 0.75% NaCl, total length, shoot length, root length, and dry weight decreased by 38.8, 34.3, 49.4, and 43.6%, respectively, relative to the control. Root length showed the greatest proportional reduction among the growth traits. The mean regression coefficients were −2.48 for total length, −2.03 for shoot length, −2.84 for root length, −2.31 for dry weight, and 3.25 for salt-injury score, revealing substantial variation among accessions. Principal component analysis of the five response coefficients retained two principal components that together explained 66.20% of the total variation. PC1 mainly represented overall growth maintenance, whereas PC2 reflected root-growth maintenance and slower progression of visible injury. Hierarchical clustering of PC1 and PC2 scores produced an exploratory three-group classification consisting of 29 relatively sensitive, 107 intermediate-response, and 74 relatively tolerant accessions. Although internal validation favored a broader two-group structure, the three-group solution was useful for describing contrasting multivariate response profiles. The reference accessions IR29 and Pokkali were assigned to the relatively sensitive and relatively tolerant groups, respectively. Several additional accessions from diverse origins and genetic backgrounds were included in the relatively tolerant group. These accessions should be regarded as putative seedling-stage salt-tolerance candidates requiring further physiological, reproductive-stage, and field validation. Overall, the combined use of regression analysis, principal component analysis, and hierarchical clustering provided a practical framework for characterizing salinity-response variation in a large rice germplasm panel.