- I am an IVADO Postdoctoral Fellow working with Dr. Guillaume Lettre at Montreal Heart Institute.
- I obtained my PhD in Quantitative Life Sciences at McGill University supervised by Dr. Josée Dupuis and Dr. Hamed Najafabadi.
- Previously, I obtained my bachelor's degree in preventive medicine and data science from Fudan University.
My research focuses on developing interpretable, efficient, and robust statistical and machine learning methods for large-scale biomedical data. My ultimate goal is to improve health for all through inclusive, reproducible, and actionable research. I have developed several efficient methods for high-dimensional correlated genetic data to improve the accuracy and efficiency of key analyses in statistical genetics:
1. To identify causal genetic variants for diseases based on statistical associations and functional annotations (fine-mapping)
- We developed SparsePro to address the key challenge of integrating statistical evidence and functional genomic data.
- Software: SparsePro
- Simulation and real data analyses: Zhang et al.
- Code to reproduce analyses: SparsePro_analysis
2. To detect shared genetic signals across different phenotypes to identify actionable targets for diseases (colocalization)
- Built upon SparsePro, we developed SharePro to account for correlation between genetic variants in colocalization analysis.
- Software: SharePro_coloc
- Simulation and real data analyses: Zhang et al.
- Code to reproduce analyses: SharePro_coloc_analysis
3. To characterize genetic effect heterogeneity across populations with different environmental exposures (GxE)
- We adapted SharePro to reduce multiple testing burden in GxE analysis.
- Software: SharePro_gxe
- Simulation and real data analyses: Zhang et al.
- Code to reproduce analyses: SharePro_gxe_analysis
- We used the core gene hypothesis to identify putative core instrument variables to reduce the risk of horizontal pleiotropy in Mendelian randomization
- Software: MR Corge
- Real data analyses: Zhang et al.
- Code to reproduce analyses: HDL->CAD
- We developed a robust version of SparsePro to account for potential inconsistencies between the GWAS population and the LD reference panel.
- Software: RSparsePro_LD
- Simulation and real data analyses: Zhang et al.
- Code to reproduce analyses: RSparsePro_LD_analysis
Check out my other projects on Google Scholar!