Known limitations

RepliSage is a deliberate simplification. Most points below are acknowledged in the paper's Discussion.

• Fork speed is constant. Each fork keeps one speed drawn from a normal distribution, whereas real speed varies along the genome and in time.
• Fork-cohesin interactions are simplified. Forks act mostly as barriers. Real encounters can be more complex: cohesive cohesin, which holds sister chromatids together, can pass through the fork, and cohesin can also be unloaded. Bypass of this kind and fork stalling are not modelled; unloading appears only indirectly, through \(c_{\text{rep}}\) and random unbinding.
• Replication fountains are not explained. Replication fountains are visible in Hi-C heatmaps, but it is not clear how they would be represented in this model, since forks act mainly as barriers here and no mechanism produces such a pattern. If fountains are your focus, RepliSage is not the right tool; see Modelling replication fountains below.
• Replication stress is simple. It is slower forks plus more origins. Fork restart and the damage response are absent.
• Abstract epigenetics. The five states do not correspond to specific marks, and how forks reorganise them is unknown. The organiser effect is an assumption that you can switch off.
• Pseudotime and \(T_{\text{mc}}\). Monte Carlo steps are not real time, and \(T_{\text{mc}}\) has no clear biological meaning. Results depend on it.
• Balanced, not fitted. Coefficients are balanced by design because no data can fit them, and the model is sensitive to them.
• Sister chromatids are symmetric. Loop extrusion and the propagation of epigenetic states act identically on both sister chromatids: the two copies share the same LEF positions and the same states, whereas in a cell each sister has its own loops and its own pattern of marks, restored independently. A more detailed model could extrude loops and propagate states separately on each sister, and let them differ.
• Coarse 3D model. Beads cover 7 to 100 kb; there is no nuclear envelope, lamina or nucleolus, and mitosis is reduced to a second LEF family, topoisomerase-like relaxation and a pulling force. Transcription is not included.

Modelling replication fountains

RepliSage is not perfect for fountains. If you want to study how DNA replication and polymer duplication reshape genome folding, including fountain-like patterns, start from the dedicated modelling work of D'Asaro and colleagues, which treats replication machineries and sister chromatids explicitly:

1. D. D'Asaro, M. M. C. Tortora, C. Vaillant, J.-M. Arbona and D. Jost, DNA replication and polymer chain duplication reshape the genome in space and time, Physical Review X 14, 041020 (2024). doi:10.1103/PhysRevX.14.041020. A minimal polymer framework with explicit replication fork progression and sister-chromatid formation; it shows that replication-dependent loop formation can shape chromosomes as loose bottle brushes during S phase.
2. D. D'Asaro, J.-M. Arbona, V. Piveteau, A. Piazza, C. Vaillant and D. Jost, Genome-wide modeling of DNA replication in space and time confirms the emergence of replication specific patterns in vivo in eukaryotes, Genome Biology 26, 431 (2025). doi:10.1186/s13059-025-03872-4. A genome-wide extension compared with in vivo data.

Trust trends more than exact numbers, and treat results that depend on the organiser assumption as hypotheses. Runs are stochastic: use --seed and --n-replicas. Found something else? Please open a GitHub issue.