Cotton is one of the most significant economic crops cultivated worldwide. Cottonseed is a strategic reservoir of high-quality plant protein and an underexploited resource for the food and feed industries. To quantify nutritional diversity and identify superior germplasm, we evaluated 312 upland cotton (Gossypium hirsutum L.) accessions over two consecutive growing seasons and characterized 30 agronomic and nutritional traits. Protein content varied widely (29.6–48.8%), with a coefficient of variation of 7.5–11.7% and a two-year mean of 37.0%. Glutamic acid (Glu; 154.0 mg/g) and aspartic acid (Asp; 90.7 mg/g) were the most abundant amino acids, and lysine and arginine were relatively high among essential amino acids. Correlation analysis based on genotype best linear unbiased estimates (BLUEs) showed that most nutritional traits were positively or neutrally associated with key yield-related traits, particularly lint percentage (LP) (e.g., protein vs. LP: r = 0.18, p < 0.01), indicating the feasibility of simultaneous improvement in seed nutritional quality and lint yield potential. Using 29 core traits with complete two-year data, we developed an integrated evaluation framework combining principal component analysis (PCA), grey relational analysis (GRA), TOPSIS, and the analytic hierarchy process (AHP) to rank accessions comprehensively. This framework identified 10 elite germplasm lines with high protein content and favorable yield potential, exemplified by “Xinluzhong 34” (Rank 1; phenotypic comprehensive value, P_i = 0.733). These results provide a quantitative foundation for value-added cottonseed utilization and support breeding strategies aimed at developing cultivars with both high yield and enhanced nutritional quality.

Cottonseed is an important resource for edible oil and protein. Here, we evaluated cottonseed oil, protein, and gossypol contents using traditional chemical methods and near-infrared reflectance spectroscopy (NIRS) in diverse upland cotton (n = 456) and sea island cotton (n = 205) germplasm. In upland cotton, oil content averaged 21.23 ± 3.98% (12.74–43.56%), protein averaged 23.63 ± 4.63% (15.53–49.79%), and gossypol averaged 1.47 ± 0.21 mg/g (0.06–2.07). Correlation analysis showed a significant negative association between oil and protein (r = −0.125, p < 0.01; confirmed by NIRS: r = −0.171, p < 0.01), a significant negative association between protein and gossypol (r = −0.375, p < 0.01), and a significant positive association between oil and gossypol (r = 0.409, p < 0.01). In sea island cotton, oil, protein, and gossypol contents averaged 24.82 ± 6.15% (14.64–41.43%), 25.75 ± 2.04% (18.84–39.00%), and 1.60 ± 0.15 mg/g (1.22–2.08), respectively. The oil–protein association was strongly negative by NIRS (r = −0.744, p < 0.01), whereas correlations with gossypol were weak and not significant by the traditional method. After screening and evaluation, high oil and protein varieties were identified in upland cotton (n = 15) and sea island cotton (n = 6). Fourteen extreme-oil upland materials were further used to examine flowering-date effects on oil accumulation and physiological indicators, showing rapid oil accumulation and a flowering-date-dependent maximum. Finally, qRT-PCR analysis of lipid-metabolism-related candidate genes showed that seven genes were expressed significantly higher in high-oil than in low-oil materials (p < 0.05), peaking at the late stage of oil accumulation. GhDGAT1 and GhDGAT2 showed positive regulatory effects on oil accumulation, whereas GhFAD3 and GhKCR2 showed negative regulatory effects. Collectively, these findings provide compositional benchmarks, clarify trait relationships, and identify candidate genes useful for breeding cotton cultivars with improved seed oil/protein traits.