FIG. 5.

Effect of model parameters on 3D reconstruction quality. We compute the 3D reconstruction as a function of parameter values using the Hi-C data matrix associated with chromosome 19 in cell 1 of the mESC data set. The top row of structures from left to right shows the effect of an increased weight λ₁ on the parameterization penalty h₁. We vary λ₁ = 0.01, 0.1, 1, 10 and fix λ₂ = λ₃ = 0 to obtain four solution curves with exponentially increasing penalty weight. The center row of structures from left to right shows the effect of an increasing weight λ₂ on the smoothing penalty h₂. Here we vary λ₂ = 0.01, 0.1, 1, 10 and fix λ₁ = 0.5 and λ₃ = 0 to obtain these four solution curves. Finally, to show the effect of incorporating the bulk data—the mESC chromosome 19 population matrix—in the analysis, we vary λ₃ = 0.01, 0.1, 1, 10 and fix λ₁ = 0.5, λ₂ = 1 to obtain the four structures on the bottom row. We computed all structures using the multiscale approach with n = 73, 146, 292, 584. 3D, three-dimensional; mESC, mouse embryonic stem cell.