Early selection of soybean yield via agronomic trait phenotyping under Asian soybean rust pressure

Abstract

Asian soybean rust (ASR) is a major threat due to its aggressiveness, fungicide tolerance, and ability to overcome resistance genes. Breeding highyielding cultivars under ASR pressure presents challenges, including low early selection accuracy and managing several lines. This study evaluated whether early agronomic traits could predict yield in later generations under ASR. Traits assessed included seed yield, days to flowering and maturity, plant height, and 50-seed weight. F2 or F2:3 data were used in regression and machine learning models to predict yield in F2:5 progenies. Similar R² values across approaches suggested a mainly linear relationship among predictors. Using F2:3 data improved R², especially for flowering, maturity, and height. Univariate models with these traits performed best, reaching R² values up to 52.12%. These models can improve early selection and reduce the breeding workload.