Integrative modeling of the genome structure and dynamics in fission yeast Authors: Soya Shinkai, Toshinori Namba, Takeshi Sugawara, Soya Hagiwara, Shuichi Onami, Tokuko Haraguchi, Yasushi Hiraoka, Akinori Awazu, Masaru Ueno, Shin-ichi Tate Published: 2026 ABSTRACT SignificanceBuilding quantitative models that connect genome structure to chromatin dynamics is essential for predicting chromosome motion and genome function in living cells. Using genome-wide live-cell imaging and Hi-C data-constrained polymer modeling in fission yeast, we build a physics-based “digital twin” of the genome consistent with spatiotemporal chromatin dynamics. This model reproduces known architectural features as validation, reveals characteristic chromatin relaxation times, and demonstrates that low-frequency forces generated outside the nucleus propagate through the spindle pole body and centromeres to drive genome-wide chromatin displacement. These results establish a physical framework for understanding how nuclear mechanics shapes genome dynamics and function, with broad implications for genome biology and biophysics. The eukaryotic genome is packaged into chromatin folded in a hierarchical order within the nucleus (1–3). If the nucleus could be observed with complete spatiotemporal resolution, it would, in principle, be possible to track the spatial and temporal aspects of genome transactions, such as gene expression, DNA replication, and DNA damage repair, and to reveal how they are linked to chromatin architecture (4, 5). In practice, however, experimental access is limited and modality-dependent. Fluorescence in situ hybridization (FISH) provides single-cell snapshots of nuclear genome organization in fixed cells, from the positioning of specific loci to chromosome-scale arrangements (6–8). In addition, chromosome conformation capture approaches coupled with high-throughput sequencing techniques, such as Hi-C (9), yield genome-wide contact maps that report ensemble-averaged three-dimensional (3D) genome features. Furthermore, fluorescent tagging of genomic loci enables live-cell imaging of chromatin motion in living cells (10, 11). Together, these approaches have established that chromatin is both structured and dynamic, and that its organization is shaped by nuclear geometry and subnuclear positioning, including constraints imposed by intranuclear structures and nuclear boundaries. However, these approaches examine different outcomes, which are often interpreted separately; therefore, it remains challenging to obtain a unified picture of chromatin as a structured yet dynamically fluctuating polymer. Therefore, a key challenge is to establish a quantitative framework that integrates chromatin structure and dynamics across modalities, enabling the interpretation of locus-resolved motion in the context of genome-wide architecture. Although live-cell imaging makes chromatin dynamics experimentally accessible, quantitative analyses largely focus on a limited number of loci, making it difficult to relate locus-specific motion to genome-wide structural features. In contrast, genome-wide Hi-C contact maps provide a comprehensive view of nuclear organization but do not directly specify how chromatin moves in time. To bridge the gap between these complementary measurements, datasets of genome dynamics comparable in scope to Hi-C and an integrative framework that can connect structural and dynamical observables are required. Such a framework would also provide a path toward predictive physics-based models that enable systematic in silico investigation of how mechanical driving forces and constraints shape genome dynamics at the nuclear scale. Fission yeast Schizosaccharomyces pombe is a genetically tractable model eukaryote whose compact genome and three chromosomes (I–III) make dense genome-wide locus tagging experimentally feasible. In this study, to address the gap between genome-wide 3D genome data and live-cell trajectories, we established a genome-wide live-cell imaging platform in S. pombe. Building on a systematic locus-tagging strategy in S. pombe, we generated a collection of fluorescently tagged strains targeting 143 genomic loci spaced at ∼90 kb intervals (12) and established 131 strains in which locus motion could be quantified together with nuclear landmarks, including the spindle pole body (SPB) associated with centromeres and the nucleolus, in the G2 phase. We obtained a comprehensive set of locus-scale dynamical measurements mapped onto genomic coordinates that could be integrated with genome-wide structural information. Next, we integrated two independent observables, genome-wide Hi-C contact data as structural constraints and locus-resolved dynamical readouts, through polymer modeling of Hi-C data, PHi-C (13, 14). Finally, we leveraged the PHi-C framework to construct a physics-based “digital twin” of the S. pombe genome: a polymer model that captures the structured and dynamic nature of chromatin, enabling in silico exploration and prediction beyond what is accessible in low-throughput experiments. Results Quantitative Mapping of Chromatin Dynamics in Fission Yeast Enabled Through Genome-Wide Live-Cell Imaging. Constructing an accurate digital twin of the genome requires large-scale measurements of chromatin dynamics at higher-resolution across many genomic loci. To systematically map chromatin motion across the S. pombe genome in living cells, we established a comprehensive live-cell imaging platform based on the lacO/LacI–GFP tagging strategy (Fig. 1A). First, we generated a collection of strains, each carrying a single lacO array integrated at one of 143 genomic positions at ∼90 kb spacing across chromosomes I–III of S. pombe, enabling locus-resolved measurements over broad genomic coverage (12) (Dataset S1). To monitor chromosome dynamics within the context of nuclear landmarks, we labeled two prominent reference structures, the SPB and the nucleolus. S. pombe centromeres cluster near the SPB and are tethered to the nuclear envelope near the SPB during interphase (15, 16), forming a characteristic Rabl configuration (SI Appendix, Fig. S1A); therefore, we tagged the SPB component Sid4 with GFP and used its focus position as a proxy for centromere motion. Simultaneously, we tagged the nucleolar protein Gar2 with mCherry to mark the nucleolus. For subsequent imaging and analysis, we used 131 strains that showed robust dual labeling. This setup enabled us to simultaneously track a single chromosomal locus (LacI–GFP), the SPB-proximal (centromere-proximal) region (Sid4–GFP), and the nucleolus (Gar2–mCherry) in the same living cells (17) (Fig. 1B and Movie S1). Comprehensive live-cell imaging of chromatin dynamics across the S. pombe genome. (A) Strategy for genome-wide fluorescent labeling of chromosomal loci. Schematic representations of the three S. pombe chromosomes are shown, with target loci (green) and centromeres. Individual cell lines were established for each locus, in which a lacO array was inserted at the target site and the loci were visualized by imaging of lacO/LacI–GFP, while Sid4–GFP marks the spindle pole body (SPB). (B) Representative fluorescence image of a living cell showing a single LacI–GFP-labeled locus and Sid4–GFP at the SPB (green), together with Gar2–mCherry, a nucleolar component showing a strong signal in the nucleolus and a weaker diffuse signal in the cytoplasm (red). (C) Typical 3D trajectories of the SPB, a labeled locus, and the nucleolus center obtained from live-cell imaging. (D) Comprehensive dataset of chromatin dynamics for 131 genomic loci in the G2 phase (3D images acquired every 30 s for 1,800 s). (E) Heat map representation of the two-point mean square displacement (MSD) between each locus and the SPB along the genomic coordinate, revealing genome-wide motion relative to the SPB. Using the aforementioned live-cell imaging setup (Fig. 1 A and B), we obtained representative 3D trajectories of the SPB, a labeled genomic locus, and the nucleolus center through time-lapse imaging (Fig. 1C). Consistent with prior observations of microtubule-dependent SPB movements in interphase (17, 18), Sid4-marked SPBs showed characteristic semiperiodic motion, which facilitated visual discrimination of SPB motion from locus trajectories, even within the same GFP channel (Fig. 1C); the frequency components of SPB motion are analyzed in detail later. We acquired 3D image stacks every 30 s for 1,800 s (61 time points) from a large collection of cells spanning multiple cell cycle stages. Therefore, we assigned cell cycle stages post hoc, and in subsequent analyses, we focused on the G2 phase, the longest stage of the S. pombe cell cycle, to characterize interphase chromatin dynamics (SI Appendix, Fig. S1B). As a result, we obtained a comprehensive G2-phase dataset of chromatin dynamics for 131 genomic loci from a total of 4,751 cells (Fig. 1D and Dataset S1). To quantify locus motion, we used this curated G2-phase dataset to calculate the two-point mean squared displacement (2-point MSD) of the 3D vector from the SPB–centromere system to each locus (Materials and Methods). A genome-wide heat map of the resulting 2-point MSDs revealed a clear trend (Fig. 1E): positional fluctuations increased with increasing distance from the centromere, consistent with progressively less constrained motion along chromosome arms. This centromere-to-arm trend is consistent with previous observations in budding yeast Saccharomyces cerevisiae, where centromere tethering confines chromatin and motion increases with distance from the centromere (19, 20). However, although this G2-phase dataset provided broad coverage of locus motion, many regions remained unmeasured (gray), leaving gaps in the genome-wide dynamic landscape. Construction of a Physics-Based Polymer Model Consistent with Genome-Wide Chromatin Dynamics. To complement the genome-wide 2-point MSD map obtained through live-cell imaging (Fig. 1E) and to infer chromatin dynamics beyond the experimentally sampled loci, we constructed a physics-based polymer model constrained with genome-wide Hi-C contact data using our PHi-C framework (13, 14). PHi-C estimates an effective interaction parameter matrix (the PHi-C matrix) for an extended bead–spring polymer model through optimization such that the resulting model reproduces the input Hi-C contact matrix (SI Appendix, Fig. S2A). We applied PHi-C genome-wide to a previously reported interphase Hi-C contact map of S. pombe (21) (>90% of the cells were in the G2 phase) at 25 kb resolution across chromosomes I–III (Fig. 2A and SI Appendix, Fig. S2B). Importantly, we applied PHi-C jointly to chromosomes I–III instead of modeling each chromosome separately. This genome-wide formulation allowed the PHi-C-inferred model to incorporate interchromosomal contacts (including clustering of the three centromeres and prominent telomere clustering among 1L, 1R, 2L, and 2R) observed in previous Hi-C experiments (22, 23). The reconstructed genome-wide contact map generated from the optimized PHi-C matrix showed excellent agreement with the experimental genome-wide Hi-C contact map (Pearson’s r=0.995; Fig. 2A and SI Appendix, Fig. S2C), indicating that the PHi-C-inferred interaction landscape faithfully captures genome-wide contact architecture. Construction of a physics-based polymer model consistent with genome-wide chromatin dynamics. (A) Comparison between the experimental Hi-C contact map and the reconstructed contact map obtained using PHi-C analysis at 25 kb resolution, showing a high correlation (r=0.995) across the three chromosomes I–III. (B) Representative fitting of theoretical 2-point MSD curves derived from the PHi-C-inferred model to experimental data for selected loci on chromosomes I–III. The 2-point MSD between the SPB and each locus is plotted against lag time. Gray curves indicate TAMSDs from individual cells, the magenta curve shows the experimental ensemble-averaged TAMSD, and the cyan curve denotes the fitted theoretical curve determined by the parameters σ (μm) and γ (kg s−1). (C and D) Fitted values of the model parameters σ (C) and γ (D) plotted along the genomic coordinates of chromosomes I–III. Each point represents one of the 131 genomic loci. The horizontal line indicates the mean value, and the shaded area denotes the ±2SD range. (E) Reconstructed genome-wide heat map of the 2-point MSD between each locus and the SPB, obtained using the PHi-C-inferred model with the mean parameter values (σ=0.0567μm and γ=1.26×10−7 kg s−1). (F) Correlation between experimental and reconstructed 2-point MSDs. Each point represents a pairwise value of the 2-point MSD using live-cell imaging and the PHi-C-inferred model, integrated over lag times from 0 s to 900 s. The probability density is estimated using Gaussian kernel smoothing, with Pearson’s correlation coefficient r=0.880. Next, to quantitatively connect the PHi-C-inferred model to chromatin dynamics, on the basis of the PHi-C matrix, we derived a theoretical expression for the 2-point MSD between the SPB (used as a proxy for the centromere) and each genomic locus; see Eq. 1 in Materials and Methods. This expression depends on two physical parameters: the effective bead diameter σ to set the spatial scale of the coarse-grained polymer and the friction coefficient γ to set the time scale. Here, σ maps bead units to physical length, whereas γ maps model dynamics to physical time, thereby anchoring the otherwise dimensionless coarse-grained polymer model to real spatiotemporal units. To estimate σ and γ, we fit the theoretical 2-point MSD curves to the experimental 2-point MSD measurements for the 131 genomic loci imaged (Fig. 2B and Movies S2–S4). Results showed that remarkably, the fitted values of σ and γ were broadly consistent across genomic positions and clustered around their genome-wide means (Fig. 2 C and D). Therefore, we adopted their mean parameter values (σ=0.0567μm and γ=1.26×10−7 kg s−1) for subsequent analyses. Using the mean values of σ and γ, we used the PHi-C-inferred model to reconstruct a genome-wide heat map of the 2-point MSD between each locus and the SPB (Fig. 2E), which closely resembled the experimental 2-point MSD map (Fig. 1E). Quantitatively, both reconstructed and experimental 2-point MSDs showed a strong correlation when integrated over lag times from 0 to 900 s (Pearson’s r=0.880; Fig. 2F). We also observed this correlation when analyzing the 2-point MSDs as a function of genomic distance from the centromere at individual lag times (SI Appendix, Fig. S3). Together, these findings demonstrate that PHi-C yields a physics-based digital twin of the S. pombe genome that is consistent with both the genome-wide Hi-C contact map and the live-cell measurements of the G2-phase chromatin dynamics. Movie S5 shows a representative simulation displaying time-evolving 3D conformations of chromosomes I–III, along with the corresponding distance map in order to visualize the chromatin dynamics generated using the digital twin model. Structural Features and Genome-Wide Mobility Patterns Derived from the Digital Twin. The digital twin model was calibrated to be consistent with the genome-wide Hi-C contact map and the live-cell chromatin dynamics; however, its agreement with independent biological observations is not guaranteed a priori. We therefore asked whether the digital twin reproduces hallmark features of S. pombe nuclear organization that were not used in building the digital twin. An independent benchmark is provided by the characteristic interchromosomal organization of interphase chromosomes, namely centromere and telomere clustering, established by previous FISH-based observations (15, 22). Exploiting the efficient sampling of 3D conformations enabled by the PHi-C framework, we generated 105 structural samples and quantified these hallmark features in a model-based genome-wide manner. To assess centromere clustering, we defined the circumradius R as the radius of the circle passing through the three centromeres (cen1–3) of chromosomes I–III (Fig. 3A, Inset). The resulting distribution of R was strongly shifted toward smaller values for centromeres (mean R=0.261μm) compared to a control distribution for three randomly selected genomic loci (mean R=0.544μm; Fig. 3A), indicating tight spatial clustering of the three centromeres in the model. Notably, the mean centromere circumradius was comparable in scale to the optical diffraction limit under typical wide-field imaging conditions, consistent with the frequent appearance of centromeres as a single cluster in imaging-based studies. Structural features and genome-wide mobility patterns derived from the digital twin. (A) Probability distributions of the circumradius R, defined as the radius of the circle passing through centromeres (cen1–3) of chromosomes I–III (Inset), computed from 105 structural samples generated using PHi-C. The distribution for the centromeres (blue) shows a smaller radius (mean R=0.261 μm) than that obtained by choosing three loci at random (orange; mean R=0.544 μm). (B) Telomere clustering patterns obtained from 105 structural samples. The four telomeres (tel1L, magenta square; tel1R, magenta triangle; tel2L, cyan square; and tel2R, cyan triangle) were grouped into clusters, with a threshold distance of 250 nm. The bar plot shows the population (%) of structures forming 1 to 4 telomere clusters. Schematic Insets illustrate representative telomere groupings for cases with two or three clusters. (C) Genome-wide heat map of chromatin mobility computed using PHi-C, showing log10MSD(t) for each 25 kb locus along chromosomes I–III (horizontal axis) across lag times from 0.1 s to 900 s (vertical axis, log scale). The colored strip beneath the heat map denotes the mobility cluster assigned to each locus: cluster 1 (red), cluster 2 (green), and cluster 3 (blue). The genomic position of the mat locus is indicated by a triangle. (D) MSD(t) profiles for the three mobility clusters identified in (C). For each cluster, MSD curves of all member loci are plotted on log–log axes, with clusters 1, 2, and 3 shown in red, green, and blue, respectively. Filled circles mark the MSDs at the relaxation time of each curve. (E) Probability distributions of relaxation times τ for the three mobility clusters. Kernel density estimate curves are shown for clusters 1, 2, and 3 in red, green, and blue, respectively. Mean relaxation times are 71.6 s, 155 s, and 157 s for clusters 1, 2, and 3, respectively. To examine telomere clustering of chromosomes I and II, we grouped the four telomeres (tel1L, tel1R, tel2L, and tel2R) into clusters using a threshold distance of 250 nm, chosen to be on the order of the optical diffraction limit (Fig. 3B). Across the 105 structural samples, configurations forming two telomere clusters were most prevalent (Fig. 3B), in agreement with previous observations (15, 22). The model further enabled quantification of the relative frequencies of distinct telomere grouping patterns beyond the overall cluster count (Fig. 3B and SI Appendix, Fig. S4), providing a more detailed, population-level characterization of telomere organization than is typically accessible using imaging alone. In summary, the digital twin model reproduces the centromere and telomere clustering of S. pombe, validating it against hallmark organizational features that were not used in building the digital twin. Having validated the digital twin against these independent structural hallmarks, we next exploited it to quantify genome-wide chromatin mobility over a broad temporal range, an analysis inaccessible to low-throughput experiments. We computed log10MSD(t) for each 25 kb locus from lag times of 0.1 s to 900 s and visualized the results as a genome-wide heat map (Fig. 3C). Notably, although the mobility clusters were defined post hoc from the full MSD(t) profiles (Fig. 3D), the heat map revealed a characteristic temporal crossover: Although loci in the centromere- and telomere-enriched loci (clusters 2 and 3) had relatively smaller MSDs for short lag times (t≲10s) compared to the remaining loci (cluster 1), they reached larger long-time MSD plateaus toward t=900s (Fig. 3C). This analysis extended the dynamic characterization beyond the experimentally sampled time scales and loci by exploiting the PHi-C-inferred model’s analytical and computational tractability. Clustering of the computed MSD(t) profiles revealed three distinct mobility classes (Fig. 3 C and D and SI Appendix, Fig. S5). Notably, the three classes were enriched for specific genomic landmarks: Telomere-proximal loci on chromosomes I and II were preferentially assigned to one cluster, centromere-proximal loci to another, and the remaining loci predominantly to the third (Fig. 3 C and D), indicating that centromeres and telomeres have distinct dynamic signatures in the model. In addition, the genome-wide MSD heat map highlighted the silent mating-type region (mat locus) on chromosome II as a locus with comparatively small short-time MSDs, similar to those observed for the centromere- and telomere-enriched clusters (Fig. 3C). To investigate this, we performed a subclustering analysis within cluster 1, which divided it into two subclasses, 1A and 1B; the mat locus was assigned to subclass 1A, whereas the majority of cluster 1 loci belonged to subclass 1B (SI Appendix, Fig. S5 B–D). In the PHi-C-inferred interaction landscape (SI Appendix, Fig. S2B), the mat locus showed comparatively strong attractive interactions with the three centromeres and the telomeres of chromosomes I and II (SI Appendix, Fig. S2D). Finally, since the computed MSD(t) curves saturated at long times, we quantified a characteristic relaxation time τ for each locus and compared τ across mobility classes. Specifically, for each MSD(t) curve, we defined the plateau value A for long times with τ as the lag time at which MSD(τ)=A(1−1/e), as shown in Fig. 3D (filled circles). The relaxation time distributions showed clear class-specific differences (Fig. 3E): The loci in the telomere- and centromere-enriched clusters took a longer time to relax (∼150 s; mean τ=155s and 157 s, respectively), whereas the remaining loci relaxed more rapidly (∼70 s; mean τ=71.6s). Together, these findings define two characteristic time scales of chromatin motion, ∼150 s for centromere-/telomere-enriched loci and ∼70 s for the remaining loci. Therefore, the digital twin provides model-grounded quantitative time scales for genome-wide chromatin mobility beyond reproducing known centromere and telomere clustering, establishing a quantitative basis for interpreting the comprehensive chromatin dynamics dataset. Semiperiodic Motion of the SPB and Frequency-Dependent Chromatin Response Derived from the Digital Twin. The relaxation time analysis (Fig. 3) revealed two characteristic chromatin mobility time scales ∼150 s and ∼70 s. Therefore, we investigated what physical processes could drive or modulate motion on those time scales in the living cells. One prominent candidate is the semiperiodic motion of the SPB (17, 18), which is mechanically coupled to clustered centromeres in S. pombe and, thus, can act as a semiperiodic driver of chromatin. Notably, in our dataset, the SPB had a mean speed of 16.8 nms−1, which is smaller than the typical 1 s scale displacement of individual nucleosomes inferred from live-cell imaging in human cells: MSD(t=1s)≈0.020μm (i.e., MSD(t=1s)∼141nm) (24). Thus, at short time scales, the directional component of SPB motion can be comparable to, or masked by, stochastic chromatin fluctuations, underscoring the importance of analyzing SPB motion over longer time windows, which is enabled by our time-lapse imaging (3D stacks acquired every 30 s for 1,800 s). Therefore, we quantified the temporal characteristics of SPB motion across the population and investigated how centromere-directed driving forces propagate through chromatin in the digital twin model. The SPB undergoes directional, semiperiodic motion (17), which we observed in our live-cell imaging dataset (Fig. 4A–C). To quantify the temporal structure of SPB motion across the population, we computed the power spectral density (PSD) of SPB trajectories for all 4,751 G2-phase cells (Fig. 4D). The PSD spectrum displayed three notable features: 1/f-like fluctuations (well approximated by 1/f0.7 for f≲1/360Hz), where f is the frequency; a local peak at f=1/225Hz; and substantial attenuation above f≈1/150Hz, where the PSD fell to about half of its peak value at f=1/225Hz (Fig. 4D). The PSD spectrum also revealed a prominent oscillatory component with a characteristic period of ∼225 s (corresponding to f≈1/225Hz), consistent with previous reports of semiperiodic SPB motion on similar time scales (17). In contrast, oscillatory components with periods shorter than ∼150 s (i.e., f≳1/150Hz) were comparatively weak, indicating that slow SPB motion dominates the nuclear mechanical driving forces that centromere-proximal chromatin experiences. Notably, these PSD features intersected naturally with the characteristic chromatin relaxation time scales identified in Fig. 3. Results showed that high-frequency SPB components at f≳1/150Hz (periods shorter than ∼150 s) are weak; therefore, SPB motion is unlikely to act as a dominant driving force input on time scales faster than the intrinsic chromatin relaxation times of the centromere-/telomere-enriched regime. Instead, the PSD spectrum suggests that SPB motion is concentrated in the low-frequency region (f≲1/150Hz), leading to the hypothesis that slow SPB motion are the mechanically relevant input for genome-scale deformations. Semiperiodic motion of the SPB and frequency-dependent chromatin response derived from the digital twin. (A) Representative fluorescence snapshot of a living cell showing the SPB (Sid4–GFP) and a LacI–GFP-labeled locus (green), along with Gar2–mCherry (red). Short-term trajectories of the two green foci highlight the motion of the SPB and the locus. (Scale bar, 0.8 μm.) (B and C) Typical 3D trajectory of the SPB in a single cell with its principal axis (B), and the corresponding time series of SPB displacement projected onto the principal axis (C), color-coded by time (30 s intervals, 61 frames). (D) Power spectral density (PSD) of SPB motion plotted against frequency f on logarithmic axes. The solid line with circle markers shows the median PSD across 4,751 cells, and the filled region indicates the interquartile range (IQR, Q1–Q3). (E) Schematic illustrating the definition of the frequency-domain linear response function when an oscillatory force is applied to the three centromeres, in which χn(ω) is decomposed into storage (χn′) and loss (χn″) components. (F) Genome-wide linear response of chromatin to oscillatory forces applied at the three centromeres. The storage (Upper) and loss (Lower) components of the response function, χ′(2πf) and χ″(2πf), are shown as heat maps across chromosomes I–III as a function of driving force frequency f (log scale). Horizontal dashed lines mark 1/225 and 1/150 Hz. (G) Heat map of the quality factor Q(f)=|χ′(2πf)/χ″(2πf)|, quantifying the efficiency with which oscillatory forces applied at the three centromeres propagate along the genome. Horizontal dashed lines mark 1/225 and 1/150 Hz. Next, we analyzed how oscillatory driving forces transmitted through the centromere–SPB linkage influence genome-wide chromatin dynamics. A key advantage of the digital twin model is that it provides a physics-based framework for computing dynamical responses, without the need for additional perturbation experiments. In the general setting, we considered applying an oscillatory force of angular frequency ω to the m-th bead and quantified the resulting displacement of the n-th bead in the frequency domain. This relation was characterized by a cross-response function χnm(ω), which could be evaluated directly using the PHi-C-inferred model parameters; see Eq. 2 in Materials and Methods and SI Appendix, Fig. S6. Following the convention in linear rheology, we obtained an expression for χnm(ω)=χnm′(ω)−iχnm″(ω), where the storage component χnm′(ω) quantifies the elastic-like energy-storing part of the response, while the loss component χnm″(ω) quantifies the viscous/liquid-like energy-dissipating part. Next, we focused on centromere-directed oscillatory driving forces. We applied an oscillatory force F(ω) to the three centromeres and quantified the resulting displacement rn(ω) of each locus n (Fig. 4E). Under the linear response relation rn(ω)=χn(ω)F(ω), we defined χn(ω) and mapped its storage and loss components, χ′(2πf) and χ″(2πf), respectively, genome-wide as a function of the oscillatory driving force frequency f (Fig. 4F). In these heat maps, the sign indicates the direction of motion relative to the applied force: Positive values (red) correspond to displacements aligned with the oscillatory force direction, whereas negative values (blue) indicate anti-aligned displacements. As shown in Fig. 4F, at low frequencies, genomic loci near the centromeres showed strong positive responses (red), whereas more distal regions tended to have pronounced negative responses (blue), indicating a global deformation mode with opposite displacements across chromosome arms. In contrast, at f≳1/150Hz, appreciable responses were largely confined to centromere-proximal loci, while the remainder of the genome responded only weakly (Fig. 4F), consistent with limited propagation of rapid forces through the chromatin polymer. Notably, at f≈1/225Hz, telomere-proximal loci on chromosomes I and II showed positive responses in both storage and loss components (i.e., χ′>0 and χ″>0), indicating coherent displacements in the oscillatory force direction with both elastic-like and dissipative contributions (Fig. 4F). To assess how efficiently oscillatory inputs propagate without dissipation, we computed the quality factor Q(f)=|χ′(2πf)/χ″(2πf)| (Fig. 4G). Although centromere-proximal loci showed the largest response amplitudes, their low Q values indicated inefficient, strongly dissipative transmission locally. In contrast, more distal regions displayed higher Q values, indicating more efficient propagation of oscillatory inputs along chromosome arms (Fig. 4G). In particular, at frequencies just below 1/225Hz, chromosomes I and II showed efficient transmission in their telomere-proximal loci, and chromosome III showed efficient responses across more than half of its length (Fig. 4G). Together, these findings indicate that the SPB–centromere system forms a mechanical driver that preferentially couples to slow, low-frequency modes of chromatin motion. An oscillatory force at f≳1/150Hz has little impact beyond the centromere-proximal region, whereas slow oscillatory inputs, including 1/f-like fluctuations and the characteristic peak component at f≈1/225Hz, can propagate genome-wide and generate large-scale nuclear displacements predicted using the digital twin model (Fig. 4 F and G). Discussion Integrating Genome-Wide Live-Cell Dynamics with Hi-C Data-Constrained Polymer Modeling. One key underlying theme of genome biology is that nonrandom spatial chromosomal organization is tightly linked to diverse DNA transactions, including transcription, replication, and genome maintenance (25, 26). Over the past decades, complementary experimental approaches have helped establish increasingly detailed but inherently partial views of nuclear organization. On the one hand, microscopy-based methods (from FISH to live-cell genomic locus tracking) provide direct access to spatiotemporal chromatin behavior. On the other hand, chromosome conformation capture (3C)-based methods (e.g., Hi-C) yield genome-wide contact landscapes but, by design, do not directly report the dynamics in living cells (27). Recent advances in sparse labeling and single-molecule and superresolution microscopy now enable tracking of chromatin motion at the nucleosome level in single cells (3, 11, 28). However, with regard to specific genomic loci, live-cell imaging has a few limitations: the number of loci that can be labeled and tracked per cell and strain construction and imaging throughput. To overcome these limitations in a genetically tractable organism, in this study, we established a genome-spanning live-cell imaging resource in fission yeast S. pombe. We constructed a collection of strains containing the markers for the SPB-proximal (centromere-proximal) region and the nucleolus, along with a lacO array inserted at one of the 131 genomic loci at ∼90 kb spacing across chromosomes I–III of S. pombe (12). Within a spatial context of interpretable nuclear landmarks as internal references, this enabled us to take systematic, locus-resolved measurements of interphase chromatin motion at a scale that is unusually large for a locus-tagging strategy. However, even such a large imaging effort samples discrete loci and time windows, leaving substantial genomic intervals unobserved and highlighting the need for a framework based on physical principles to infer genome dynamics beyond the sampled loci and to enable mechanistic interpretation. In this study, we addressed this gap by integrating the imaging-based chromatin dynamics map with a genome-wide Hi-C contact map through a physics-based polymer model inferred using our PHi-C framework (13, 14). A key advantage of this approach is that it yields an explicit physical model of chromosomes in which stochastic fluctuations arise naturally and can be analyzed quantitatively. Therefore, the physics-based digital twin is inherently dynamic, enabling direct computation of chromatin motion in addition to genome architecture. We calibrated the model’s physical parameters directly against the experimental 2-point MSDs, simultaneously constraining the interaction landscape with a genome-wide Hi-C contact map, and obtained a dynamical polymer model that is consistent with both genome-wide contacts and the observed chromatin mobility across the genome. Notably, performing PHi-C jointly across chromosomes I–III allowed interchromosomal constraints present in the Hi-C map (including centromere clustering and telomere-associated interactions) to be incorporated from the beginning, enabling a unified description of nuclear-scale organization and dynamics. This integrated model enables the computation of chromatin motion beyond the experimentally sampled loci and the connection of genome-wide mobility patterns to underlying physical constraints in the nucleus. Interpreting Genome-Wide MSD Profiles Through a Dynamical Polymer Digital Twin. By calibrating the PHi-C-inferred model to the experimentally measured 2-point MSDs, we obtained a dynamical representation of the S. pombe genome that is consistent with both genome architecture and dynamics in the G2 phase. In this sense, the model functions as a physics-based digital twin by providing a quantitative framework for inferring chromatin behavior beyond the genomic loci and time scales that are directly accessible using locus-tagging live-cell imaging. One advantage of this digital twin view is that it enables mechanistic, population-level readouts that are difficult to obtain experimentally. For example, structural sampling allows quantitative decomposition of telomere clustering into distinct grouping patterns (Fig. 3 and SI Appendix, Fig. S4). In addition, calibrated dynamics allows genome-wide computation of locus-specific mobility over a broad temporal range (subseconds to many minutes; Fig. 3C). Importantly, going beyond qualitative descriptions, these computations translate full MSD(t) profiles into interpretable dynamical quantities. Clustering of the computed MSD(t) curves revealed three mobility classes that map naturally onto nuclear landmarks (Fig. 3 C and D and SI Appendix, Fig. S5). Centromere- and telomere-proximal loci form two distinct classes (clusters 2 and 3) characterized by a temporal crossover in which short-time mobility is relatively suppressed (t≲10s) but long-time movements are larger compared to other loci (Fig. 3C). Each MSD(t) curve was saturated for long times in the PHi-C model, so we further extracted a characteristic relaxation time τ as a quantitative time scale for nuclear fluctuations. Notably, the relaxation time distributions could be separated into two regimes: centromere- and telomere-enriched loci relaxed within ∼150 s, whereas the remaining loci relaxed faster within ∼70 s (Fig. 3E). Furthermore, although the classes were defined purely from the computed MSD(t) profiles, constitutive heterochromatin domains at centromeres and telomeres (29) were preferentially assigned to cluster 2 and 3, whereas the remaining loci across chromosome arms were predominantly assigned to cluster 1 and appeared more euchromatin-like. The silent mating-type region (mat locus), another constitutive heterochromatin domain positioned at the nuclear envelope (29–32), further clarifies this concept. The mat locus shares the reduced short-time mobility characteristic of heterochromatin-enriched clusters, forming a distinct subclass within cluster 1 (SI Appendix, Fig. S5 B–D); however, its relaxation time is closer to that of the faster regime (τ=80.5s; Dataset S2). In summary, these findings are consistent with the condition under which the mat locus experiences short-time confinement, while relaxing more rapidly for longer times. One possible interpretation is that short-time confinement primarily reflects a local chromatin state, whereas long-time relaxation is additionally shaped by the strength and persistence of nuclear landmark associations. Therefore, centromere and telomere heterochromatin may relax slowly because of stronger or more persistent tethering; in contrast, the mat locus retains heterochromatin-like short-time confinement yet relaxes more rapidly in the case of comparatively weaker or more transient nuclear landmark associations. Such locus-specific combinations of short-time confinement and long-time relaxation may be relevant to how heterochromatin maintains stable yet regulatable epigenetic states. In particular, studies have proposed that perinuclear positioning of heterochromatin suppresses histone turnover and promotes epigenetic inheritance (33); the distinctive long-time relaxation behavior of the mat locus raises the possibility that this locus might tune the stability–plasticity balance by modulating its nuclear positioning and associated histone turnover. These characteristic time scales raise the question of what cellular processes provide mechanical inputs on comparable time scales and how such inputs propagate through the genome. Therefore, we next examined the semiperiodic motion of the SPB and used the digital twin model to quantify the frequency-dependent chromatin response to centromere-directed oscillatory driving forces (Fig. 4). SPB Oscillations as a Geometry-Tuned Mechanical Driver of Genome-Wide Chromatin Mobility. Previous live-cell imaging studies have established the interphase motion of the SPB in S. pombe (18, 34–38). These studies have also simultaneously visualized SPB-associated structures and microtubule dynamics, revealing microtubule-dependent SPB/nuclear movements in living cells. In interphase cells, the nucleus is oscillated by the force balance between cytoplasmic microtubules with opposite directions (18). A more striking example of SPB motion is observed in meiotic prophase of S. pombe (34, 35). Meiotic prophase SPB-led nuclear oscillations (horsetail movement) are a hallmark of fission yeast and are driven by cytoplasmic microtubules (35, 36). These results have provided a canonical mechanism for how microtubule-generated forces can reorganize nuclear contents over time in interphase and meiotic prophase (37, 38). Beyond these descriptive observations, several studies have also investigated how cytoskeleton-driven motion shapes chromatin mobility and function. In fission yeast, centromeres pulled by cytoplasmic microtubules facilitate the mobility of associated interphase genomic regions, such that microtubule destabilization reduces the mobility not only of centromeres but also of these loci (39). Loss of the microtubule nucleator Mto1 reduces microtubule-dependent chromosomal movement consistently and is associated with impaired DNA repair phenotypes and altered sister-chromatid cohesion (40). In budding yeast S. cerevisiae, cytoplasmic microtubules constrain chromatin motion while actin filaments drive nuclear rotation, and actin depolymerization lowered the subdiffusive exponent of the SPB and a tracked locus, whereas microtubule depolymerization released the centromere and increased global chromatin mobility (41). However, despite these important advances, the physical principles governing how such cytoskeleton-driven forces propagate through the chromatin polymer to mobilize the genome remain largely unexplored using quantitative modeling. It is also unclear which time/frequency components are most effective at deforming the genome. One exception is that SPB motion exhibits a characteristic oscillatory peak with a period of 203 to 229 s under specific metabolic conditions (17), which is consistent with the pronounced peak near f=1/225Hz and the strong contribution of low-frequency components that we observed in the PSD spectrum in this study (Fig. 4D). Therefore, a key goal is to characterize the genome as a mechanically coupled system driven by SPB–centromere-associated motion and to identify the frequency range in which such oscillatory driving forces can generate large-scale chromatin displacements. To this end, we leveraged the PHi-C digital twin to compute, on the basis of physical principles, how a force applied at one locus is transmitted as displacement at other loci (SI Appendix, Fig. S6). In the general setting, this transmission is described using a cross-response function χnm(ω), as shown in Eq. 2 in Materials and Methods; we focused specifically on centromere-directed forcing by applying an oscillatory force to the three clustered centromeres and mapping the resulting storage and loss components of the genome-wide response (Fig. 4E–G). Results revealed a significant frequency dependence: High-frequency forces at f≳1/150Hz are largely confined to centromere-proximal loci, whereas low-frequency driving forces propagate more effectively across chromosome arms (Fig. 4 F and G). Together with the relaxation time analysis (Fig. 3), these findings imply that SPB motion is unlikely to be a dominant driver of chromatin fluctuations on the relaxation time scales of ∼70 s and ∼150 s. Instead, SPB motion might act primarily through slower, low-frequency mechanical inputs that can propagate over chromosome arms. Notably, at approximately f=1/225Hz, telomere-proximal loci on chromosomes I and II exhibit positive responses in both storage and loss components (i.e., χ′>0 and χ″>0, respectively), indicating coherent displacements aligned with the centromere-directed forces in both components (Fig. 4F). Consistent with this view, the PSD spectrum shows substantial power in the low-frequency band, with a 1/f0.7 profile (Fig. 4D), indicating that SPB motion contains broad, slow components rather than being dominated by a single period. Although the origin of this scaling is beyond the scope of this study, its practical implication is clear: Broadband low-frequency driving forces fall precisely in the regime that our response analysis predicts to transmit most effectively through the genome (Fig. 4 F and G). Such low-frequency driving forces may therefore shape genome-wide chromatin dynamics not only through a peak component but also by broadening the spectrum of accessible chromosome fluctuations, rather than selectively exciting a single periodic response. Together, these findings show that our quantitative, frequency-domain framework can interpret how SPB motion can be coupled to genome-wide chromatin dynamics. Importantly, the dominant frequencies of SPB motion may not be universal but can be shaped by cell geometry and microtubule dynamics. In interphase S. pombe cells, SPB motion is mechanically linked to microtubule polymerization dynamics that interact with cell ends, indicating that the effective oscillation period depends on cell size and shape (18, 42). Also in meiotic prophase, nuclear oscillations show clear size dependence that can be captured using a force balance model relating microtubule-generated forces to viscous drag (43). These observations support the idea that fission yeast may exploit its rod-shaped geometry to tune SPB-driven mechanical inputs toward low-frequency components, which our response analysis predicts to be most effective at propagating genome-wide. Finally, the ability of SPB-coupled low-frequency driving forces to mobilize chromatin has potential functional implications. In fission yeast, reduced microtubule-dependent chromosomal motion is correlated with compromised DNA repair phenotypes (40), consistent with a model in which large-scale nuclear fluctuations contribute to maintaining genome integrity. More broadly, although diverse eukaryotic systems show increased chromatin mobility in response to DNA damage (44–46), the mechanical route by which such mobility is generated is likely shaped by organism-specific nuclear geometry and chromosome architecture. In this regard, the rod-shaped cell morphology and the small number of relatively long chromosomes in S. pombe may favor an SPB-driven, low-frequency mechanical strategy that differs from the mechanisms that organisms with more isotropic cell shapes and/or many shorter chromosomes use. Summary and Outlook. In this study, we combined comprehensive live-cell locus tracking with Hi-C data-constrained polymer modeling to build a dynamical, physics-based digital twin of the S. pombe genome. Genome-wide calibration of the PHi-C-inferred model against experimental 2-point MSDs enabled quantitative computation of chromatin mobility beyond the experimentally sampled loci and time scales; simultaneously, it preserved the contact architecture across chromosomes I–III. This integrated model reveals nuclear landmark-associated dynamical regimes with characteristic relaxation time scales of ∼70 s and ∼150 s and helps obtain a frequency-domain description of how SPB–centromere-directed periodic driving forces propagate through the genome, highlighting the dominant role of low-frequency mechanical inputs in generating large-scale nuclear deformations. The strategy of integrating an imaging-based genome-wide dynamics map with a genome-wide Hi-C contact map through PHi-C is in principle applicable to other organisms. For substantially larger genomes, however, applying this strategy will require addressing both the achievable coverage of locus labeling and the increasing computational cost of PHi-C modeling as the matrix size grows. Hi-C contact maps implicitly reflect constraints imposed by nuclear boundaries (e.g., nuclear envelope attachments) and by subnuclear compartments (e.g., the nucleolus); therefore, such models offer a route to interpret genome organization and dynamics at the level of the nucleus as a physical system, rather than chromosomes in isolation. Moreover, the frequency response analysis provides a general framework to incorporate externally imposed or internally generated forces and to connect mechanical driving forces to functional outputs. In general, a systems-level understanding spanning the extracellular mechanical environment, nuclear organization, and genome regulation is becoming increasingly important. Extracellular matrix stiffness can modulate transcriptional programs via mechanosensitive genomic enhancers (47). In this context, tuning the period and amplitude of SPB-driven motion, for example, by altering the cell size or mechanical conditions, might systematically reshape frequency-dependent chromatin responses derived from the digital twin model. An important next step will be to determine which biological functions are coupled to such changes in chromatin responses to SPB oscillations, such as genome maintenance and gene regulation. More broadly, building physically grounded chromosome models will continue to provide a way to uncover the physical significance of chromatin dynamics within the nucleus and the cell as an integrated system. Materials and Methods Strain Construction. S. pombe strains (h90 ade6-210 leu1 ura4-D18 lys1 XXX[::ura4+-KanR-lacOp] his7+::lacI-GFP), carrying a single lacO array inserted at one of the 143 genomic loci (XXX), were used to introduce Sid4–GFP and Gar2–mCherry as described previously (17), yielding 131 strains with robust dual labeling (Dataset S1). Briefly, the Sid4–GFP–nat plasmid was linearized using StuI and used to transform the S. pombe strains. Subsequently, the Gar2–mCherry–hph plasmid was linearized using AflII and used to transform the resulting S. pombe strains. Live-Cell Imaging. Cells were grown in YEA medium (0.5% yeast extract, 3% glucose and 40 μgm−1L adenine) at 30 °C overnight. A total of 2.5×105 cells from a log-phase culture were added to a glass-bottom dish [Iwaki; coated with 5 mgm−1L of lectin from Bandeiraea simplicifolia BS-I (Sigma)] for live-cell imaging. A stage-top incubator equipped with a lens heater (Tokai Hit) was used to maintain the sample temperature at 30 °C. Each sample was used for a maximum of 3 h of imaging before cells reached a stationary phase. Microscopic images of asynchronous cells were obtained using an iXon3 897 EMCCD camera (Andor) connected to a Yokogawa CSU-W1 spinning-disc scan head (Yokogawa Electric Corporation) and an Olympus IX83 microscope (Olympus) with a UPlanSApo 100× NA 1.4 objective lens (Olympus). Laser illumination was performed at a wavelength of 488 nm for Sid4–GFP and LacI–GFP, and at 561 nm for Gar2–mCherry. Images were captured using MetaMorph software (Molecular Devices). Optical section data (41 focal planes with 0.2 μm spacing, acquired every 30 s; 160 nm/pixel) were collected for 30 min, as shown previously, and we found that the cells could still divide after imaging, suggesting that this imaging condition does not damage cells (48). Image Processing and 3D Spot Tracking. Time-lapse sequences were deconvolved using Huygens image analysis software (Scientific Volume Imaging). All deconvolved 3D time-lapse image stacks (STK format), along with the corresponding ND metadata files, have been deposited in the SSBD:repository (ssbd-repos-000493). The dataset is organized by the 131 genomic locus IDs (Dataset S1) and the date of imaging. A video index was assigned as a prefix to each filename. Within each imaging file, individual cells were annotated with a unique cell index, and a cell cycle index based on the morphological criteria shown in SI Appendix, Fig. S1B was manually assigned: 1, G1/S; 2, early G2; 3, G2; 4, late G2; and 5, other. These annotations were combined to define the final cell ID used in subsequent analyses. The deconvolved time-lapse sequences for each locus were analyzed using IMARIS software (Bitplane). For Sid4–GFP and LacI–GFP, the center of each fluorescent focus was tracked in 3D over 30 min at 30 s intervals for each strain. For Gar2–mCherry, the geometric center of the strong fluorescence region was tracked as the 3D position of the nucleolus center, as described previously (17). The resulting G2-phase trajectories consist of SPB (X,Y,Z), locus (X,Y,Z), and nucleolus-center (X,Y,Z) coordinates recorded over 61 frames (61×9 values), along with the corresponding cell cycle index; all data are available in the SSBD:repository (ssbd-repos-000493). Two-Point Mean Square Displacement Between the SPB and the Locus. For each cell, 3D trajectories of the SPB and a LacI–GFP-labeled locus were represented as rSPB(tj)j=060 and rlocus(tj)j=060 (tj=jΔt, Δt=30 s). To remove nuclear translation, we computed the relative position vector as slocus(tj)=rlocus(tj)−rSPB(tj). For each relative trajectory slocus(tj)j=060, we calculated the time-averaged mean square displacement (TAMSD) for lag times τk=kΔt(k=0,1,2,⋯,30) using the following equation:TAMSDlocus(τk)=161−k∑j=060−kslocus(tj+τk)−slocus(tj)2. The ensemble average of TAMSDlocus(τk) over all tracked cells was used as the 2-point MSD between the SPB and the locus, and the MSDs were visualized genome-wide as a heat map (Fig. 1E). PHi-C Analysis. Hi-C preprocessing. Hi-C FASTQ data for wild-type interphase S. pombe cells (SRR5227728, SRR5227738, SRR5227739, and SRR5227749) obtained from Kakui et al. (21) were processed following the 4DN Hi-C scripts (https://github.com/4dn-dcic/docker-4dn-hic/tree/v43/scripts). Hi-C reads were mapped to the S. pombe reference genome (ASM294v2) using bwa (v0.7.18), and alignments were handled using samtools (v1.21). Valid Hi-C contacts were identified using pairtools (v1.1.0) with the subcommands parse, sort, dedup, split, and select, followed by indexing using pairix (v0.3.8). The four resulting pairs files (.pairs.gz) were merged, and a genome-wide Hi-C file (.hic) was generated using juicer_tools (v1.22.01) with the pre command (MAPQ ≥ 30 with genome-wide Knight–Ruiz GW_KR normalization). Optimization of the PHi-C matrix. A contact matrix for PHi-C analysis was generated from the .hic file using hic-straw (v1.3.1). Intra- and interchromosomal contacts were extracted at 25 kb resolution using GW_KR normalization, producing a single square matrix whose size corresponded to the total number of 25 kb genomic bins across the three S. pombe chromosomes: chromosome I (5,579,133 bp → 224 bins), chromosome II (4,539,804 bp → 182 bins), and chromosome III (2,452,883 bp → 99 bins); this yielded a 505×505 matrix. To remove poorly mappable genomic regions, such as centromeres, we excluded rows and columns in which ≥60% of entries were NaN, resulting in a dense 501×501 contact matrix that was used as input for PHi-C optimization. Dataset S2 lists the genomic coordinates corresponding to indices n=0 to 504, including those removed during filtering, with centromeric indices highlighted. PHi-C optimization was performed using phic (v2.0.13), with the preprocessing and optimization subcommands executed sequentially. The optimized interaction matrix (PHi-C matrix) reproduced the input Hi-C contact map with high fidelity, achieving a Pearson correlation of r=0.995 between the experimental and reconstructed contact maps (SI Appendix, Fig. S2). Theoretical computation of the 2-point MSD from the PHi-C-inferred model. We computed the theoretical 2-point MSD on the basis of the PHi-C-inferred model following the formalism described by Shinkai et al. (13) From the optimized N×N PHi-C matrix, we constructed the corresponding normalized Laplacian matrix L¯ and diagonalized it to obtain the normalized eigenvalues 0=λ¯0<λ¯1<⋯<λ¯N−1 and the orthogonal matrix Q. Let M denote the set of indices corresponding to centromeric bins. In the model, the centroid of the beads belonging to set M is treated as the position of the SPB. Under these assumptions, the theoretical 2-point MSD between the SPB and the n-th locus can be computed analytically as follows:[1]MSDlocus(t;n,cen)=2σ2∑p=1N−1Qnp−1|M|∑m∈MQmp2λ¯p−11−e−3kBTγσ2λ¯pt, where σ is the effective polymer segment size (contact distance parameter), γ is the friction coefficient of each bead in the PHi-C-inferred model, kB is the Boltzmann constant, and T is the temperature (K). Fitting of model parameters (σ and γ) to the experimental 2-point MSD. For each of the 131 genomic loci, the experimentally measured 2-point MSD between the SPB and the locus was fitted to the theoretical 2-point MSD function derived from the PHi-C-inferred model (see the previous section). We evaluated the theoretical 2-point MSD at T=303K, corresponding to the live-cell imaging conditions. Next, nonlinear least-squares fitting was performed using scipy.optimize.curve_fit with the following settings: initial parameter values σ=0.05μm and γ=1×10−7 kg s−1, and bounds = (0, ∞) for both parameters. The two parameters σ and γ were simultaneously estimated for each locus. Structural sampling from the optimized PHi-C-inferred model. Structural sampling was performed using the sampling subcommand in phic (v2.0.13), generating 105 3D configurations of chromosomes I–III of S. pombe. Next, the sampled coordinates were scaled using the experimentally determined value σ=0.0567μm to match the actual physical length scale measured in live-cell imaging. For each sampled structure, we defined the centromere position of each chromosome as the centroid of the bead coordinates corresponding to the centromeric indices in the PHi-C-inferred model. Using the three centromere coordinates (for chromosomes I–III), as well as triplets of beads randomly selected from each chromosome model, we computed the circumradius. For three points A, B, and C in 3D space, with side lengths of a=|BC|, b=|CA|, and c=|AB| and triangle area S, we calculated the circumradius R from the geometric relation R=abc4S. In addition, we extracted the four telomeres of chromosomes I and II (tel1L, tel1R, tel2L, and tel2R) from each sampled structure and quantified telomere clustering by grouping telomeres whose pairwise distances fell below a threshold. The proportions of structures forming 1 to 4 telomere clusters were computed for threshold distances of 200 nm, 250 nm, and 300 nm (SI Appendix, Fig. S4). Genome-wide chromatin mobility analysis. Using the genome-wide mean values of the fitted parameters obtained from experimental 2-point MSDs (σ=0.0567μm and γ=1.26×10−7 kg s−1), we computed the dynamic properties of the three-chromosome physics-based polymer model generated using PHi-C. For each genomic index n in the PHi-C-inferred model, the single-locus mean square displacement MSD(t;n) was calculated using the following theoretical expression (13, 49):MSD(t;n)=2σ2∑p=1N−1Qnp2λ¯p−11−e−3kBTγσ2λ¯pt, where the center-of-mass mode (p=0) was removed. To classify the shapes of the MSD(t) curves, we computed the cosine similarity between all pairs of MSD profiles evaluated on a log–log scale using cosine_similarity from sklearn.metrics.pairwise. Next, we constructed a cosine distance matrix as 1−(cosine similarity), with diagonal elements set to zero. Finally, we performed hierarchical clustering on this distance matrix using scipy.cluster.hierarchy.linkage with method="average", visualized the resulting dendrogram using scipy.cluster.hierarchy.dendrogram, and defined three clusters by cutting the dendrogram (SI Appendix, Fig. S5). Power Spectral Analysis of SPB Motion. We conducted power spectral analysis of SPB motion using the SPB and nucleolus center coordinates obtained from 3D time-lapse imaging (Δt=30 s, 61 frames). To remove the drift arising during image acquisition, we expressed the SPB coordinates relative to the nucleolus center at each time point. For each degree of freedom (x, y, and z), we constructed a one-dimensional time series, and computed the PSD using the scipy.signal.welch function with parameters fs = 1/30 and nperseg = 60. Finally, we summed the PSDs of the three degrees of freedom to obtain the total PSD for each SPB trajectory. This analysis was applied to all G2-phase cells, yielding PSD profiles for 4,751 cells. Frequency-Domain Response Function Analysis. The dynamics of the polymer model consisting of N segments in PHi-C are described by the Langevin equationγdr(t)dt=−Lr(t)+ξ(t), where r=(r0,r1,⋯,rN−1) and ξ=(ξ0,ξ1,⋯,ξN−1) denote the position and thermal noise supervectors, respectively (13, 50). The matrix L is the Laplacian matrix derived from the interaction matrix of the PHi-C-inferred model. When an external force fm(t) is applied to the m-th segment, the Langevin equation can be written using the unit vector em (whose m-th component is 1 and all others are 0) asγdr(t)dt=−Lr(t)+ξ(t)+fm(t)em. Since the Laplacian matrix L derived from the PHi-C interaction matrix is a real symmetric matrix, it admits the spectral decomposition L=QΛQT. Multiplying from the left by QT and defining the normal coordinates X=QTr, we obtain:γX˙(t)=−ΛX(t)+QTξ(t)+fmQTem. Since Λ=diag(λ0,λ1,⋯,λN−1), the equation of motion decouples into independent equations for each mode p:γX˙p(t)=−λpXp(t)+ηp(t)+Qmpfm(t), where we used η(t)=QTξ(t) and (QTem)p=Qmp. Applying the Fourier transform ∫−∞∞dte−iωt yields the following relations between the Fourier components:iγωXp(ω)=−λpXp(ω)+ηp(ω)+Qmpfm(ω).∴Xp(ω)=ηp(ω)+Qmpfm(ω)λp+iγω. Since the inverse transformation is given by r=QX, the Fourier component of the displacement of the n-th segment becomes rn(ω)=∑p=0N−1QnpXp(ω). Taking the ensemble average of both sides and using ⟨ηp(ω)⟩=0, the linear response relationship of the n-th segment to an external force applied to the m-th segment in the frequency domain is written as rn(ω)=χnm(ω)fm(ω). Thus, we obtain the cross-response function as[2]χnm(ω)=χnm′(ω)−iχnm″(ω)=∑p=0N−1QnpQmpλp+iγω. Therefore, the real and imaginary parts of χnm(ω) define the storage and loss response functions, respectively, given byχnm′(ω)=∑p=0N−1QnpQmpλp2+γ2ω2λpandχnm″(ω)=∑p=0N−1QnpQmpλp2+γ2ω2γω. Using the normalized eigenvalues λ¯p obtained from the PHi-C analysis, we computed the physical eigenvalues as λp=3kBTσ2λ¯p. To focus on the propagation of forces within the S. pombe three-chromosome system, we removed the contribution of the center-of-mass mode (p=0). Next, we calculated the response function of the n-th segment to an oscillatory force applied to the centromeres, χn(ω) (Fig. 4E), by averaging the cross-response functions over the set M of centromeric indices:χn(ω)=1|M|∑m∈Mχnm(ω). Supplementary Material Appendix 01 (PDF) Dataset S01 (CSV) Dataset S02 (CSV) Live-cell imaging of the chromosomal locus and nuclear landmarks in S. pombe. Time-lapse sequence corresponding to Fig. 1B. A cell expressing Sid4–GFP and LacI–GFP (green), along with Gar2–mCherry (red), was imaged at a time interval of 30 s for 61 frames (total duration 1800 s). Sid4–GFP labels the SPB, LacI–GFP marks a single genomic locus, and Gar2–mCherry shows a strong nucleolar signal with weaker diffuse cytoplasmic fluorescence. Scale bar, 0.7 μm. PHi-C fitting of 2-point MSDs for all loci on chromosome I. Time-lapse visualization corresponding to the representative fitting shown in Fig. 2B. For each of the 62 loci on chromosome I, grey curves show TAMSDs from individual cells, the magenta curve shows the experimental ensemble-averaged TAMSD, and cyan curve represents the fitted theoretical MSD curve derived from the PHi-C-inferred model. PHi-C fitting of 2-point MSDs for all loci on chromosome II. Time-lapse visualization of the fitting process for the 44 loci on chromosome II. Grey curves denote TAMSDs from individual cells, the magenta curve shows the ensemble-averaged TAMSD, and the cyan curve represents the corresponding theoretical MSD curve based on the PHi-C-inferred model. PHi-C fitting of 2-point MSDs for all loci on chromosome III. Time-lapse visualization of the fitting process for the 25 loci on chromosome III. Grey curves represent TAMSDs from individual cells, the magenta curve shows the ensemble-averaged TAMSD, and the cyan curve shows the fitted theoretical MSD curve derived from the PHi-C-inferred model. Dynamical behavior of the digital twin model. A simulation demonstrating the chromatin dynamics generated using the digital twin model. (Left) Time-evolving 3D configurations of chromosomes I, II, and III (magenta, cyan, and orange, respectively) are shown, with centromeres marked by bead representations and telomeres indicated by yellow beads. (Right) Corresponding pairwise distance map updated at each frame. The elapsed time (0–5 s) is displayed above the distance map.