Seminar by Prof. Zhonghua LIU from Department of Biostatistics, Columbia University

Hidden confounding bias is a major threat in identifying causal protein biomarkers for Alzheimer’s disease in non-randomized studies. Mendelian randomization (MR) framework holds the promise of removing such hidden confounding bias by leveraging protein quantitative trait loci (pQTL) as instrumental variables (IVs) for establishing causal relationships. However, some pQTLs might violate core IV assumptions, leading to biased causal inference and misleading scientific conclusions. To address this urgent challenge, we propose a novel MR method called MR-SPI that first Selects valid pQTL IVs under the Anna Karenina Principle and then performs valid Post-selection Inference that is robust to possible pQTL selection error. We further develop a computationally efficient pipeline by integrating MR-SPI and AlphaFold3 to automatically identify causal protein biomarkers and predict protein 3D structural alterations. We apply this pipeline to analyze genome-wide summary statistics for 912 plasma proteins in 54,306 participants from UK Biobank and for Alzheimer’s disease (AD) in 455,258 samples. We identified seven proteins associated with Alzheimer's disease - TREM2, PILRB, PILRA, EPHA1, CD33, RET, and CD55 - whose 3D structures are altered by missense genetic variations. This discovery offers novel perspectives into their biological roles in AD development and may aid in identifying potential drug targets.

Professor LIU Zhonghua is Assistant Professor in the Department of Biostatistics at Columbia University since 2022. His primary research interests include statistical genetics/genomics, causal inference, and machine/deep learning algorithms and their applications. He was Assistant Professor in the Department of Statistics and Actuarial Science at The University of Hong Kong from 2018 to 2022, and he worked at Morgan Stanley in New York City from 2016 to 2018. He obtained his doctorate in biostatistics from Harvard University in 2015 and received postdoctoral training at Harvard University and Broad Institute of Harvard and MIT from 2015-2016.