AUTHOR=Araújo Maurício S. , Pavan João P. S. , Stella André A. , Fregonezi Bruno F. , Lima Natally F. , Leles Erica P. , Santos Michelle F. , Goldsmith Peter , Chigeza Godfree , Diers Brian W. , Pinheiro José B. 

TITLE=Optimizing soybean variety selection for the Pan-African Trial network using factor analytic models and envirotyping

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2025.1594736

DOI=10.3389/fpls.2025.1594736

ISSN=1664-462X

ABSTRACT=Soybean is a global food and industrial crop, however, climate change significantly affects its grain yield. Therefore, the selection of varieties with high adaptation to target population of environments is imperative in Sub-Saharan Africa. This study aimed to identify soybean varieties with high overall performance and stability using multi-environment trial data from the Pan-African Soybean Trial Network. Additionally, we sought to determine the environmental factors influencing yield through envirotyping tools. In two South-Eastern African countries, a total of 169 soybean varieties were evaluated across 83 environments in 19 locations in Malawi (47 trials) and 14 locations in Zambia (36 trials). The trials followed a randomized complete block design with three replications. Data for 37 environmental features were obtained from NASA POWER and SoilGrids. We fitted factor analytic models (FA) to estimate genotype adaptation across environments. Additionally, we applied an environmental kernel approach and the XGBoost method to assess the number of mega-environments. The FA model with four factors provided the best fit, explaining 82.44% and 81.95% of the variance and the average semi-variance ratio (ASVR), respectively. Approximately, 59.6% of the genotype-by-environment interaction were crossover. Varieties V025, V035, and V158 exhibited high yield potential and reliability but displayed moderate stability. Three mega-environments were identified, with growing degree days, mean temperature, and photosynthetically active radiation use efficiency being the most associated features for soybean grain yield. To enhance the identification of variety adaptation in these environments, integrating machine learning models with crop growth modeling is essential to assess associations between environmental features and soybean yield.