Prediction of peanut protein solubility based on the evaluation model established by supervised principal component regression

- There was intimate correlation between protein solubility and other indexes.
- At 0.05 level, a model with 11 indexes was established by SPCR.
- At 0.01 level, another model with 4 indexes was established.
- The former model was more accurate and the latter was more convenient.

Supervised principal component regression (SPCR) analysis was adopted to establish the evaluation model of peanut protein solubility. Sixty-six peanut varieties were analysed in the present study. Results showed there was intimate correlation between protein solubility and other indexes. At 0.05 level, these 11 indexes, namely crude fat, crude protein, total sugar, cystine, arginine, conarachin I, 37.5 kDa, 23.5 kDa, 15.5 kDa, protein extraction rate, and kernel ratio, were correlated with protein solubility and were extracted to for establishing the SPCR model. At 0.01 level, a simper model was built between the four indexes (crude protein, cystine, conarachin I, and 15.5 kDa) and protein solubility. Verification results showed that the coefficients between theoretical and experimental values were 0.815 (p < 0.05) and 0.699 (p < 0.01), respectively, which indicated both models can forecast the protein solubility effectively. The application of models was more convenient and efficient than traditional determination method.