Genomes in 3D: connecting structure and function
How are micron-long chromosomes spatially organized by molecular interactions between proteins at the nanometer scale? Acting as a molecular microscope, genome-wide chromosome conformation capture (Hi-C) reveals that genomes are intricately folded in 3D. Here I describe how biophysical simulations and machine learning approaches enable interpretation of these large-scale genomic datasets. First, I describe converging theoretical and experimental evidence arguing that Cohesin-mediated loop extrusion with CTCF-defined barriers plays a crucial role in interphase. Second, I describe how convolutional neural networks enable accurate predictions of genome folding from DNA sequence alone. Together, these advance our understanding of the proteins driving and the sequences underpinning 3D genome folding.