Deciphering chromatin rearrangements in response to UV irradiation using new deep learning tools for cryo-electron tomography data analysis – DEEPNER

Irradiation with UV light results in bulky DNA lesions that are repaired by Nucleotide Excision Repair (NER) pathway. NER deficiency is linked to Xeroderma Pigmentosum (XP), a human disease characterized by a high rate of skin cancers. Bulky lesion repair in vivo requires chromatin reorganization to allow access of damage recognition and repair factors. The structural aspects of this process have remained obscured because of the the low spatial resolution of imaging techniques.
Combining in situ cryo-electron tomography (cryo-ET) and deep learning based denoising we recently obtained three-dimensional imaging data that allow manual annotation of nucleosome and DNA linkers directly in the context of cell nucleus.
To take full advantage of the resolution of cryo-ET and to efficiently extract information for NER analysis, we propose a transdisciplinary project DEEPNER that brings together specialists in molecular and cell biology, cryo-ET, computational biology and image processing. Using genome editing tools, we will obtain cell lines that will be synchronously paused at given stages of the NER reaction for subsequent cryo-ET data collection. We will develop original methods and deep learning-based algorithms for the automated analysis of cryo-ET data at two structural scales: (i) identification of nucleosomes and analysis of their spatial distribution; (ii) analysis of nucleosome conformations. The focus will be on developing weakly supervised approaches that will be robust to noise and distortions observed in cryo-ET images. This project will allow us to determine the structural aspects of chromatin reorganization that are critical for ensuring successful NER in vivo. The developed software tools will facilitate the analysis of other challenging sub-cellular systems, particularly ones with inherent structural heterogeneity, such as phase-separated condensates.