SLAI Seminar 28th Session will be discussing the topic on "Chromatin Folding: Physical Modeling and AI Prediction",  from 10am to 11:30am, March 10th (Tuesday) at Room B401, online participation is welcome (Tencent Meeting ID: 734-891-927)

About the Speaker:

Dr. Huang Kai graduated with a bachelor's degree from the Department of Engineering Physics at Tsinghua University in 2009 and obtained his Ph.D. in Materials Science from the University of Wisconsin-Madison in 2015. He subsequently conducted postdoctoral research in the Department of Biomedical Engineering at Northwestern University. In 2020, he joined Shenzhen Bay Laboratory as a junior principal investigator, dedicating his work to interdisciplinary fields such as four-dimensional genomics, biological phase separation, and AI4Biology, as well as the development of related theoretical and computational methods. Dr. Huang has received the Annual Postdoctoral Award for Biophysics from the Biophysical Society in the United States and Pearl River Talents Program of Guangdong Province Young Top Talent Project. His research, as first and corresponding author (including co-authorship), has been published in renowned academic journals such as Nature, Cell Research, Science Advances, Nature Communications, and JACS. His work on chromatin higher-order structures was described by Francis Collins, former director of the Human Genome Project, as "a new image of the human three-dimensional genome." His research has also received support from projects such as the National Natural Science Foundation of China's Innovative Research Group (Category B) and general programs.

Abstract: Chromatin, as the physical carrier of genetic information, its dynamic three-dimensional structure regulates key life processes such as gene expression and DNA replication. To provide a unified explanation for the long-range interactions, density heterogeneity, and self-similarity observed in chromatin higher-order structures, we proposed the "3D Forest Model of Chromatin," predicting the widespread existence of multi-site higher-order interactions at the single-cell level. Recent Pore-C experiments have validated the universality of chromatin multivalent interactions and revealed that certain functional regulatory elements exhibit significant interaction synergy. To identify key genomic loci involved in multivalent interactions, we integrated Hi-C and epigenetic data to construct and train a graph neural network model, enabling the AI-based prediction of chromatin multivalent interactions. Further analysis indicated that distal cis-regulatory elements participate in gene expression regulation through multivalent interactions. We experimentally validated the AI-predicted synergistic effects of enhancers using CRISPRi technology, supporting the biological validity of the model. This study demonstrates the potential of artificial intelligence in predicting genome structure and function, while providing a robust theoretical framework and computational tool for research in four-dimensional genomics.