Each human cell packs over two metres of DNA into a nucleus a few microns across. The compaction is remarkable, but it is also functional: a neuron and a skin cell carry the same genome yet express very different sets of genes. A large part of that difference lies in how the DNA is folded, which stretches are buried and which are left accessible to transcriptional machinery. Understanding how physicochemical principles give rise to specific structures and the functional outcomes thereof is a central problem in genome biology.
The material that packages the genome is chromatin, a hierarchical polymer whose repeating unit is the nucleosome: ~150 base pairs of strongly negatively charged DNA wrapped around a positively charged protein core, with disordered, cationic protein tails protruding from each one. Physically, chromatin is a highly charged, multivalent heteropolymer, and its organisation is inherently multiscale: the chemistry of individual residues sets the interactions between nucleosomes, and through them the structure and function of chromatin on the scale of kilobases, a separation of scales that has been difficult to bridge in simulation.
In this talk I will describe a new residue-resolution coarse-grained chromatin model that spans this gap by simulating hundreds of nucleosomes while explicitly representing every DNA base pair and amino acid, reaching system sizes an order of magnitude beyond existing methods at this resolution. Using this model, I will show how DNA mechanics and histone chemistry shape the structure of individual chromatin fibres, and how the properties of these fibres govern collective behaviour, including phase separation and the material properties of the condensed phase.
Simulation and experiment provide complementary views of this problem. Cryo-electron tomography can image chromatin inside condensates and cells, but resolves nucleosome positions far better than the connecting DNA or histone tails. Our current work combines machine-learning-based tracing of chromatin fibres through tomograms with simulations restrained by the resulting nucleosome positions and connectivity, reconstructing complete structural models from the images. In the longer term, we aim to bring this computational toolkit together with these experimental methods to understand how chromatin organisation regulates transcription.
