A comprehensive investigation of Genotype-Environment interaction effects on seed cotton yield contributing traits in Gossypium hirsutum L. Using multivariate analysis and artificial neural network

The prediction of seed cotton yield is a critical aspect of cotton breeding. In the present study, an artificial neural network (ANN) and a multiple linear regression (MLR) model were used to predict seed cotton yield based on experimental data obtained from quantitative traits measured under different environmental conditions, including number of bolls per plant (NB), boll weight (BW), 100 seed weight (SI), number of sympodia per plant (NS), lint index (LI), internode length (IL), and seed cotton yield per plant (SCY). The experimental data underwent ANOVA and correlation analysis across different environments. The selected features were utilized for ANN and MLR modeling. The results demonstrated that the ANN model provided precise predictions of SCY, with a root mean square error (RMSE) of 6.63 g/plant and a determination coefficient (R²) of 0.888, which outperformed the MLR model, which showed an RMSE of 8.613 g/plant and an R² of 0.816. Sensitivity analysis revealed that the number of bolls per plant had the most significant impact on yield estimates, while 100 seed weight had the least impact, as determined by both ANN and MLR models. Furthermore, the ANN model was less influenced by environmental factors than the MLR model, as indicated by R² values. Overall, this study provides a comprehensive analysis of genotype-environment interaction effects on seed cotton yield and contributes to the development of effective cotton breeding strategies.