Li Meinhold Advances AI Tools for Wheat Breeding

At the Bay Area Plant Hub meeting hosted by University of California, Berkeley, Li Mainhold, a PhD student from the Small Grains Crop Breeding Program presented new research on using artificial intelligence to improve crop breeding.

Li Meinhold is developing a prediction model using WheatCAP data from public wheat breeding programs. The model combines genomic data—about 500,000 DNA markers from ~8,400 wheat lines—with 40 environmental variables such as weather and location. 

The model uses a neural network to capture how genes and environment interact to affect yield. It was tested using several scenarios, including predicting performance in new years, environments, and varieties. Results show improved accuracy when environmental data are included, though prediction remains challenging in new conditions. 

We are now refining the model for individual programs, such as UC Davis. Adding data from CIMMYT may further improve accuracy.