Figure 4: Simulations of affinity-based compartmentalization and activity-based loop extrusion recapitulate Hi-C contact patterns.

(A) Hi-C map and compartment annotations highlighting the three distinct features of S compartments. Data correspond to Chr22: 11,625,000–18,435,000 from PD-D2 sample. (B) Minimal models of A, B, and S compartmentalization with A-A and B-B affinities (both set to 0.075 kT) and S-S and heterotypic interactions set to 0.00 kT. In silico Hi-C and observed-over-expected Hi-C maps for models of (left) compartments only; (middle) compartments plus high extrusion S; (right) compartments plus high extrusion in S and low extrusion in A and B. (C) Evidence of loop extrusion in the PD-D2 Hi-C data including insulated domains, stripes, and corner peaks. Three regions of the PD-D2 Hi-C maps at 5 kb resolution, with compartment and gene tracks below. The regions correspond to (left) Chr10: 10,640,000–11,900,000; (middle) Chr11c: 7,900,000–8,615,000; (right) Chr11c: 14,310,000–14,820,000. (D) (Top) The contact probability P(s) curve for PD-D2 Hi-C as a function of genomic distance (s). (Bottom) Log-derivative of contact probability as a function of genomic distance. Curves represent averages of either chromosome-wide (gray) or for contiguous segments of a given compartment type (colored). Estimates of the average loop size in S (ℓ_S) and in B (ℓ_B) are noted along the x-axis. (E) General workflow for comparing the experimentally generated Hi-C to the polymer models of Chr15: 0–6,500,000. (F) Observed-over-expected values of S-S_inter plotted as a function of separation between loop extruders in S (d_S) for a series of polymer models. Each curve represents a series of models, which share loop extruder processivity. All models lacked extrusion in A and B chromatin and shared homotypic attractions of 0.12, 0.04, and 0.00 kT for A, B, and S, respectively, to yield an A/B compartment strength similar to the experimentally generated Hi-C. The experimental value of S-S_inter is shown as a purple line. (G) Observed-over-expected values of S_intra plotted as a function of relative abundance of loop extruders (d_AB/d_S) for a series of polymer models, each with different processivities and separations. The experimental value of S_intra is shown as a purple line. (H) Best polymer model versus the experimental data for Chr15: 0–6,500,000. The best model’s parameters are A-A attraction energy = 0.16 kT, B-B attraction energy = 0.08 kT, S-S attraction energy = 0.00 kT, λ = 55 kb, d_S = 19 kb, d_AB = 190 kb. (Left) Experimentally generated Hi-C (top, right half of the map) versus the in silico generated Hi-C map from the best model (bottom, left half of the map). (Right) Summary statistics (observed-over-expected off-diagonal compartment and on diagonal S_intra enrichments) for the experimentally generated data (left) and best model (right). (I) Summary of successful polymer models. Parameter values for (center) extruder properties and (right) homotypic A-A and B-B affinities. Models were sorted by (left) the mean square error of their observed-over-expected summary statistics versus those of the experimental data (top-most parameter set corresponds to the best model visualized in Figure 4H).