Using a deep learning model to predict and select disease-resistant plants

As the world’s population grows, agriculture must develop to meet our increasing food production needs.  In order to meet this challenge, Dr. Yongle Li is developing the tools to improve the plant breeding process with deep learning models and advanced DNA sequencing technologies. View Halo Profile >>

Tell us about your research 

The focus of my research is to develop molecular breeding tools to assist plant breeding. Conventional plant breeding process is time consuming and resource-intensive. Marker-assisted selection and genomic selection can shorten this process and save cost significantly. My group uses advanced DNA sequencing technologies and statistical models (mixed-linear model and machine learning) to identify DNA markers and genes associated with disease and yield-related traits in crops. However, complex traits such as grain yield are hard to predict, as they are often controlled by a large number of small effect genes with complicated interactions. Genomic selection is a new breeding tool for predicting and selecting complex traits of an individual plant from high-density DNA markers spread across the entire genome.

Conventional plant breeding process is time consuming and resource-intensive. Marker-assisted selection and genomic selection can shorten this process and save cost significantly.

Can you explain that to a non-scientist?

Why are some individual plants  more resistant to diseases or produce more grain? This is determined by genetics, environments and their interactions. DNA sequencing is a powerful tool for understanding the genetic mechanism of important traits and translating this knowledge into selection tools for breeding purposes.

Why did you choose this area of research?

I am always fascinated by the genetic mechanisms underlying plant development and animal behaviour. I like research that can solve “real” problems and has an impact on people’s lives. Plant breeding is an old discipline undergoing rapid rejuvenation with the help of advanced technologies such as DNA sequencing technology.

How could your Grants4Ag project someday impact #healthforall #hungerfornone?

Agriculture faces many challenges in the 21st century, including diseases outbreaks, climate change and drought. To feed an estimated 9.1 billion people by 2050, overall food production needs to increase by 70 percent. There is an urgent need to use modern breeding technologies to develop crop varieties with important traits such as disease resistance, climate resilience, high yield, etc.

There is an urgent need to use modern breeding technologies to develop crop varieties with important traits such as disease resistance, climate resilience, high yield, etc.

In this Grants4Ag project, my group will use a deep learning model, a sub-area of machine learning, to predict and select disease-resistant plants by combining major genes and minor genes to produce crops with durable disease resistance. Genomic selection using deep learning models could maintain and even increase genetic diversity in crops and be part of an integrated pest management strategy.