Polymer modelling predicts gene structural heterogeneity and transcription
Understanding the connection between genome folding and function has been a long-standing goal in genome biology. In particular, deciphering the causal relation between gene structure and transcription is a highly scrutinised question within this topic; however, the answer to this is still elusive as it remains challenging to probe chromatin structure and transcription dynamics concurrently in experiments. Here, we shed light on this problem by means of large-scale simulations using the highly predictive heteromorphic polymer (HiP-HoP) model, a fitting-free model without training to existing Hi-C data. Our simulations predict a panoply of 3D conformations of all genes across the human genome and their transcriptional activity, which highly corroborate existing experimental data. By mining this predicted pan-genomic data set, we unravel generic biophysical principles linking gene structure and expression. We find that gene transcription activity is strongly associated with the microphase-separation of DNA regulatory elements, and cell-to-cell variation in activity, or transcriptional noise, is linked to heterogeneity in gene structure. Our work thus offers structural mechanisms driving transcription and its noise, which could be relevant to cell development and differentiation.