Fig. 1: The TADdyn key steps for simulating 4D dynamic changes in a locus.

The shown example corresponds to the reprogramming dynamics for the Sox2 locus from B cells to PSC. a Data collection. In situ normalized Hi–C interaction matrices³⁵ for the region of 1.5 Mb centered around the Sox2 promoter. b Distance restraints definition. Both LowerBound- (blue) and Harmonic (red) spatial restraints are obtained by filtering the Hi–C interaction maps using the optimal triplet of TADbit parameters (“Methods”). c TADdyn steered dynamics runs. The simulated regions are represented as polymers made of spherical particles each a 5 kb-bin of the input Hi–C matrix. Particles are colored from red (first particle) to blue (last particle of the modeled region). The TSS of the locus of interest is represented by a black particle and the promoter by green ones. The models for all the stages (that is, from B to PSC) were dynamically built by TADdyn. d Contact maps (dcutoff = 200 nm) of models during the simulation were used to assess their accuracy by means of its Spearman correlation coefficients (SCCs) with the Hi–C input matrices. The SCCs of the contact maps with the seven Hi–C input matrices are shown with the coefficient in bold black letters corresponding to the time point of the column. As our previous modelling benchmark indicates⁴¹, all the SCCs > 0.60, that are obtained between models and Hi–C maps at corresponding cell stages, are indicative of good models. In c, d, only the models and the contact maps at the correspondent experimental time-points are shown, but TADdyn allows to visualize the entire dynamics filling the blanks between stages as shown in the Supplementary Movies 1–21.