Collaborative Research: DMS/NIGMS 1: Understanding genomic organization in the nucleus via biophysical modeling, super-resolution imaging, and data-efficient operator learning

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

Addressing the questions of how the two-meter-long human DNA fits into the space of a cell's nucleus (~20 um) and how it is organized within this space has been among the major mysteries of cell biology. DNA is packaged into the nucleus in the form of chromatin, consisting of a complex between DNA and histone proteins. DNA wrapped in compacted histones is thought of as “repressed” and “inaccessible”, and thus chromatin compaction plays a critical role in regulating gene activity. Current chromatin modeling is based on polymer simulations at different levels of resolution. However, given the slow time scales of these processes of the order of minutes to hours, the size scales of the order of 5--10 um (typical size of nucleus) are not accessible using methods such as molecular or dissipative dynamics approaches. The objective of this project is to decode the quantitative relationship between the physical microenvironment, multiscale 3D genome organization, and transcriptional output. The project will employ a convergent research strategy that integrates super-resolution microscopy, genomics, biophysical modeling and simulation, and machine learning. The new tools developed in this project will impact many areas in biology, including normal and abnormal tissue development, tissue degeneration in disease, as well as tissue regeneration. The research team will educate future scientists and a diverse workforce with a collaborative expertise in interdisciplinary training. Additional outreach activities will include research experiences for undergraduate students and high school students. Tissue-resident cells continuously sense changes in their chemo-physical environment and use this information to maintain their phenotype and tissue homeostasis. The project will develop a predictive framework of emergent epigenetic and transcriptional features of cells in response to changes in their physical environment. The project will develop new quantitative models for the distribution of heterochromatin domains in the interior of the nucleus as well as along the nuclear periphery. Specifically, a mathematical model will be developed to study the effect of rates of histone tail acetylation, methylation, and transcription on determining the distribution of heterochromatin domains in the interior of the nucleus. The project will further extend this model to include the formation of lamina-associated domains (LADs) by incorporating the energetic interactions between chromatin and the nuclear lamina via chromatin anchoring proteins. To verify the theoretical model, cells of fixed fate and fluid fate will be grown under varying micro-environments, and their whole genome organization at the nano- and micro-scale will be visualized and quantified through super-resolution microscopy. In addition, a machine learning framework leveraging novel deep neural operators will be developed for nonlinear inverse problems to extract the high-dimensional parameter fields implicitly and explicitly from noisy experimental images. To enhance the robustness, accuracy, and efficiency of neural operators with small data, the project will endow neural operators with prior knowledge, physics, multifidelity data, and active learning. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.