Supplementary data: In Silico ODE Modeling Reveals TP63 and TWIST1 Dosage Compensation Mediated by microRNA-Transcription Factor Networks in Aneuploid Cancer
Year: 2026
Type: supplementary-materials
Abstract: Supplementary Figure Supplementary Figure S1.pdf: Result of the fitting of the ODE models in COPASI. (A) No significant differences were observed between the experimental data and the final model predictions, indicating a successful fit. (B) Each heatmap shows experimental measurements (left), model-simulated values (center), and their absolute differences (right) across multiple cell lines (x-axis) and species (y-axis). (C) The networks of molecular interactions associated with the same models are shown. Solid and dashed lines represent upregulation and downregulation, respectively. Supplementary Figure S2.pdf: Results of multi-parameter scan on the kinetic parameter values for the species in the TP63 minimal model (A-G) and the TWIST1 (H-K) model under two copy numbers of TP63 or TWIST1 (1 and 3). The ratio between both scans is used to characterize the landscape of gene dose compensation and to analyze its dependence on the kinetic parameter values of the species. Cases of gene dosage compensation are located within the compensated region (cyan and blue colors). The “Control fold change” corresponds to the result obtained with the original fitted model. Combinations with an undefined ratio, as well as those for which COPASI did not find a steady state in one or both conditions, appear as unfilled gaps. These results indicate that the presence of a loop motif is not sufficient to generate the system-level property of gene dosage compensation. On the contrary, this property depends on the combination of specific parameter values. Supplementary Figure S3.pdf: Results of the identifiably analyses for the species in the TP63 minimal model (A) and the TWIST1 (B), obtained using the Profile Likelihood tool implemented in COPASI according to the Sabber (2012). Each graph shows the value of the objective function by successively setting one parameter and re-optimizing the remaining ones. The red curve represents the objective function value. The blue line represents the 95% confidence interval. The red line does not cross the blue line at both ends in any case, indicating that none of the parameters is identifiable. This result is consistent with the expected from the model equations and the single-point-time measurements of NCI-60 panel, which represents a steady-state experimental design. For example, in the steady state the derivatives of the species are zero, so the concentration depends on the ratio ks/kd. Without time series data, these parameters cannot be distinguished individually, although their values affect the temporal dynamics. Therefore, changes in one of these parameters in the identifiability analysis can be offset by a change in the other parameter. Other pairs (or groups) of parameters are expected to have similar behaviors, such as kr in TP63 together with ks in miR-155. Supplementary Figure S4.pdf: Comparison of experimental and predicted values by species in the HNF4A (A), TP63 (B), TP63 minimal (C) and TWIST1 (D) models. For each case, the root mean square error (RMSE), mean absolute error (MAE), coefficient of determination (R²), and Pearson's r are presented. Many factors can contribute to discrepancies between simulated and experimental data, such as the high heterogeneity among cancer cell lines, including differences in tissue origin, aneuploidy pattern, mutation, and distinct expression values at identical copy number, as well as the non-inclusion of a large number of molecules that interact with the species in the model. This is specifically relevant because each cell line is simulated using a shared set of kinetic parameters. The main value of these models lies in providing a mechanistic framework for studying gene dosage compensation, rather than predicting exact concentrations in individual cancer cell lines. Supplementary Figure S5.pdf: Base-10 logarithm transformation for the number of copies (x-axis) and the expression fold change (y-axis) for the target TF of the fitted models. Note that a slope (m) below 1 indicates sublinear growth, an m close to 1 indicates linear growth, and an m above 1 indicates superlinear growth. TP63 and TWIST1 clearly show sublinear growth. * Reduced model obtained after removing the species whose inhibition had no effect on TP63 levels. COPASI Models and Experimental Data Each model is provided as a separate .tar archive containing the COPASI model (.cps) and its corresponding experimental data (Experiments.tab) extracted from the NCI-60 cell line panel. HNF4a_model.tar: Contains the fitted ODE model of the HNF4A regulatory network (HNF4a_model.cps) and its experimental data (Experiments.tab). TP63_initial_model.tar: Contains the initial fitted ODE model of the TP63 regulatory network (TP63_initial_model.cps) and its experimental data (Experiments.tab). TP63_minimal_model.tar: Contains the reduced or minimal TP63 model (TP63_minimal_model.cps), used to identify the molecular components responsible for dosage compensation, and its experimental data (Experiments.tab). TWIST1_model.tar: Contains the fitted ODE model of the TWIST1 regulatory network (TWIST1_model.cps) and its experimental data (Experiments.tab). External Resources The regulatory interactions, CNV data, mRNA expression, and microRNA expression datasets from the NCI-60 and CCLE panels, as well as the BioNetUCR software used to construct regulatory networks and generate ODE models, can be found in Acón et al. (2021). References: Schaber, J. (2012). Easy parameter identifiability analysis with COPASI. Biosystems, 110(3), 183-185. M. Acón et al., “Myc dosage compensation is mediated by miRNA transcription factor interactions in aneuploid cancer,” iScience, vol. 24, p. 103407, Dec. 2021.
Institutions: Universidad de Costa Rica, Fidélitas University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7221618733