​

Research Interests:

Discovery of genetic variation that underlies quantitative traits is central to the genetic improvement of crops. Investigating the flow of information from genetic variation to phenotypic variation for important agronomic traits is complicated (and also facilitated) by two facts: 1) several intermediate phenotypes (e.g. mRNA and protein) are lying between genome sequences and agronomic traits; 2) most natural variations exist at low frequencies, with their effects difficult to tackle by conventional methods such as association mapping.

I am interested in studying these intermediate phenotypes by using deep learning tools, such as convolutional neural networks, to make predictions along the central dogma. I have developed deep learning models that took as input genomic regions flanking the coding sequences, and predicted the expression patterns of corresponding genes at the transcriptional or translational level. I have also dissected these models with several gradient-based or perturbation-based methods to identify important cis-elements that control gene expression. These models can be used to predict the effects of natural variations regardless of their frequencies in the population, and are thus potentially applicable in genomic prediction and maize breeding.

​

Education:

•    Ph.D. in Plant Biology, Fudan University, Shanghai, China

•    B.S. in Biology, Fudan University, Shanghai, China