Connecting Chromatin Structures to Gene Regulation Using Dynamic Polymer Simulations.

The transfer of regulatory information between distal loci on chromatin is thought to involve physical proximity, but key biophysical features of these contacts remain unclear. For instance, it is unknown how close and for how long two loci need to be in order to productively interact. The main challenge is that it is currently impossible to measure chromatin dynamics with high spatiotemporal resolution at scale. Polymer simulations provide an accessible and rigorous way to test biophysical models of chromatin regulation, yet there is a lack of simple and general methods for extracting the values of model parameters. Here we adapt the Nelder-Mead simplex optimization algorithm to select the best polymer model matching a given Hi-C dataset, using the MYC locus as an example. The model's biophysical parameters predict a compartmental rearrangement of the MYC locus in leukemia, which we validate with single-cell measurements. Leveraging trajectories predicted by the model, we find that loci with similar Hi-C contact frequencies can exhibit widely different contact dynamics. Interestingly, the frequency of productive interactions between loci exhibits a non-linear relationship with their Hi-C contact frequency when we enforce a specific capture radius and contact duration. These observations are consistent with recent experimental observations and suggest that the dynamic ensemble of chromatin configurations, rather than average contact matrices, is required to fully predict long-range chromatin interactions.

A universal deep-learning model for zinc finger design enables transcription factor reprogramming.

Cys2His2 zinc finger (ZF) domains engineered to bind specific target sequences in the genome provide an effective strategy for programmable regulation of gene expression, with many potential therapeutic applications. However, the structurally intricate engagement of ZF domains with DNA has made their design challenging. Here we describe the screening of 49 billion protein-DNA interactions and the development of a deep-learning model, ZFDesign, that solves ZF design for any genomic target. ZFDesign is a modern machine learning method that models global and target-specific differences induced by a range of library environments and specifically takes into account compatibility of neighboring fingers using a novel hierarchical transformer architecture. We demonstrate the versatility of designed ZFs as nucleases as well as activators and repressors by seamless reprogramming of human transcription factors. These factors could be used to upregulate an allele of haploinsufficiency, downregulate a gain-of-function mutation or test the consequence of regulation of a single gene as opposed to the many genes that a transcription factor would normally influence.