List of all simulation parameters and their values.

Simulation setup. A) Optimization principle. Contact frequency maps calculated from the simulated structures are compared to Hi-C data using Spearman Correlation; the simulations are then iteratively optimized to get the best-fit parameter set. B) Nelder-Mead algorithm flowchart. C and D) Loop extrusion and domain interaction models (parameters are listed in Table 1). E,G,I) Contact frequency maps of ground truth simulated data (top) compared to contact frequency maps of best-fit models (bottom). F,H,J) Comparison of the biophysical parameter values between ground truth and best-fit models. K) Model of a locus rearranged in CTCFmut (chr3: 59,800,000-61,840,000). Each TAD is assigned a specific Eattr value. The intervening TAD boundaries are each assigned two directional boundary permeabilities (possibilities of stopping at the boundary in either direction, stallL and stallR). L) Top: Hi-C contact frequency maps in WT and CTCFmut. Bottom, best-fit simulations for WT and CTCFmut. M) Best-fit biophysical parameters of the simulations in L. **p < 0.05.

Simulation of the rearrangement of the MYC locus in T-ALL. A) Locus Model: five TADs surrounding the MYC gene are assigned distinct Eattr1-5, while the four intervening boundaries are characterized by directional permeabilities (StallL1-4, StallR1-4). Boundary 3 (arrow) is disrupted in CUTLL1 and T-ALL patients. ChIP-seq and Hi-C map are plotted from a previous publication(16) (chr8:126,720,000-131,680,000). The log value of the contact matrix is plotted. B) Hi-C (top) and simulated (bottom) contact frequency maps of the MYC locus, using a polymer model featuring loop extrusion alone (no Eattr). The log values of the contact matrices are plotted. C) Hi-C (top) and simulated (bottom) contact frequency maps of the MYC locus using the combined model with both loop extrusion and Eattr. D) Best-fit parameter sets of the simulations in C. Left, TAD boundary permeability; Right, Eattr. The result shows significant differences in TAD2 and TAD4 Eattr. E) Correlations between Hi-C data and simulated contact frequency maps with boundary permeability of the disrupted boundary varying from 0 to 1 in 0.25 steps while other parameters are set to the best-fit of CUTLL1. T cell best-fit with high Eattr best correlates with the Hi-C data. *p<0.1, **p < 0.05, ***p < 0.01.

Simulations recapitulate the MYC locus disruption measured by DNA FISH. A) Top: Simulated Hi-C data; bottom: Simulated DNA FISH results; Center: CUTLL1; Left: simulated T cell with additional Eattr compared to CUTLL1; Right: simulated data with an additional loop extrusion boundary compared to CUTLL1. B) Top: Representative snapshots of simulated MYC locus structures of T cells (left) and CUTLL1 (right) with TAD2 and TAD4 marked in orange and cyan respectively; bottom: representative FISH images of T cells (left) and CUTLL1 (right) using probes tiling TAD2 and TAD4 respectively. C) Cumulative probability distributions of the distances between the centroids of TAD2 and TAD4 in CUTLL1 and naive T cells obtained in simulations and in DNA FISH experiments. Shading represents standard deviations of each bin in the cumulative probability distributions (n = 3).

Simulated chromatin dynamics. A) Contact frequency definition. On-times are defined as time intervals during which two loci are closer than the set capture radius. B) Frequency of contact (<400 nm) with the promoter (monomer 811, green) as a function of genomic position across the locus in simulated data (top) and CUTLL1 Hi-C data (bottom). Two loci of interest (monomers 859, 1590) exhibit similar contact frequency despite very different distances to the promoter. C) Example simulated time traces of the distance between the promoter and either locus marked in B. D) Left: LEF knock-down/overexpression is modeled by enforcing different LEF densities in the simulation. Center, Right: Average On- and Off-time durations of contacts (<200 nm) between the promoter and monomers separated by different genomic distances, for different LEF densities. Each data point is the average of 10 independent simulations. E) Distance-gated productive interactions occur when a locus is within a set capture radius from the promoter. F) Simulated frequency of productive interactions with the promoter (located at the distance origin) as a function of genomic position across the locus, for different capture radii. G) Frequency of productive interactions as a function of the simulated Hi-C contact frequency, for different capture radii (Hi-C capture radius: 200 nm; minimum duration set to 15s for all curves; color scheme matches that of F, from light to dark: 100 nm, 300 nm, 400 nm, 500 nm) H) Duration-gated productive interactions occur when a locus is within a set capture radius from the promoter for at least a minimum duration. I) Simulated frequency of productive interactions with the promoter (located at the distance origin) as a function of genomic position across the locus, for different minimum durations. J) Frequency of productive interactions as a function of the simulated Hi-C contact frequency, for different minimum durations (Hi-C minimum duration: 15 s; contact radius set to 200 nm for all curves; color scheme matches that of I, from light to dark: 1 min, 2 min, 5 min). K) Frequency of productive interactions as a function of Hi-C contact frequency when enforcing time- and distance-gating; (productive interaction capture radius and minimum duration for each curve are indicated in the box; Hi-C capture radius and minimum duration: 200 nm, 15 s). L) Average On- and Off-time durations of productive interactions (capture radius: 400 nm; minimum duration: 2 min) between the promoter and monomers separated by different genomic distances, for different LEF densities. Each data point is the average of 10 independent simulations.