Figures and data
A – Chromosomal region modeled in this study showing insertion sites of the MS2 (enhancer, blue rectangle) and the parS (promoter, green pentagon) from the seven fly lines of [38]. The compartments C0/C1 into which they fall were computed from the contact map in B using eigenvector analysis (Methods). B – Hi-C contact map of chromosome 2R (left, from [48]) with the region modeled in the present study enlarged (right). The enhancer and promoter sites use the same markers as in A. C – Schematic of the computation of the two observables from [38] used to assess consistency with experimental results. On the left, the mean spatial distance between loci R(s) and on the right, the two-locus mean squared displacement MSD2(t). D Schematic of an ideal chain, i.e., a bead-spring polymer chain without excluded volume. For illustrative purposes, the beads corresponding to the enhancer and promoter site are marked in blue and green, respectively. E – Comparison between the Hi-C maps from the experiment and an ideal chain simulation. F – Fit for R(s) between the ideal chain simulation and the experiment. Theoretical behavior for an ideal chain indicated by the black dashed line. Gray dotted line shows the linear fit to experimental data (slope β = 0.31) and the orange dotted line to that of the simulation (β = 0.49) G – Fit for MSD2(t) for the seven values of genomic distance s between the simulation and the experiment. The black dashed line shows theoretical power law behavior
A – Positions of LEFs from the lattice-based stochastic simulation (springs marked in red) are integrated into the bead-spring polymer chain with excluded volume (Methods). LEF activity is controlled by the parameters pU, pL, and pS, which correspond to LEF unloading, loading, and stepping probabilities, respectively. B – Mean-squared radius of gyration 
A – (Left) Bead-spring block copolymer chain with excluded volume and two bead types colored based on the compartment in which they belong. The beads corresponding to the enhancer and promoter site are marked as before. (Right) The short-range attractive potential between beads of the same type is given by the parabolic well of depth ε (in red), while the repulsive interactions enforce excluded volume (in blue) for all beads. Interactions between beads of different types are only repulsive. B – Radius of gyration of each enhancer-promoter segment as a function of compartmentalization strength ε. Values of ε where best fits with experimental data occur (from following inset with the experimental data) are indicated like in 2. C – σHi-C, σstr, and σdyn as a function of ε. The position of occurrence of best fits is marked in red. D – comparison of Hi-C contact maps between simulation and experiment for the best fit to Hi-C. E – best fit with experimental R(s) with dotted lines indicating linear fits (β = 0.52 for the simulated data). The black dashed line shows ideal chain slope. F – best fit with experimental MSD2(t). Black dashed line indicated same as before. G – relaxation time computed from the best-fit simulation to R(s) and MSD2(t). Dashed line indicates scaling exponent τ ~ 0.88
A – Schematic of loop extrusion on a compartment-based block copolymer. ε, pL, pU, and pS are the same as indicated previously. B grid with sum of normalized deviations 
Parameters of loop extrusion for stochastic simulations on the lattice
Fits for the Rouse model with excluded volume (i.e., a swollen coil) A – Hi-C, B – ⟨R⟩, C – MSD2
A – Low contact threshold k = 4, best fit at ε = 0.285. B – High threshold k = 6, best fit at ε = 0.275. C – Choosing a suitable cut-off from comparisons with Hi-C data for the block copolymer simulations.
A – Grid search with parameter pairs of ε and ⟨Ω⟩ for the compartment-based block copolymer with loop extrusion showing σstr, two minima highlighted. B – grid showing σdyn for each simulation. Minima highlighted as in inset A. C – σHi-C as a function of key parameters
Summary of deviations for all observables: Hi-C (1 - Spearman’s correlation), ⟨R⟩ (absolute deviation in logarithmic scale), and MSD2 (same as before) across all explored models
Effect of localization error on end-to-end distance ⟨R⟩ is negligible for all simulations, A – Rouse model, B – Best-fit loop extrusion, C – Best-fit compartment-based block copolymer, D – Global best-fit for loop extrusion on compartment-based block copolymer
