Spatial transcriptomics has transformed modern biology by preserving the physical address of every measured cell. Platforms such as Xenium, CosMx SMI, MERFISH, Visium HD and Stereo-seq can now profile hundreds to thousands of genes in millions of cells positioned within a tissue. Yet the sheer scale of these datasets has created an analytical dilemma. Most existing computational methods force researchers to fix their analytical lens at one resolution, typically the single cell, and to treat every cell as an independent unit. Biology, however, does not operate at a single scale. Disease processes may only become visible when dozens of interacting cells are considered together, while tissue-level architecture emerges only at even coarser resolutions.

SpatialZoomer addresses this problem with a mathematical strategy borrowed from graph signal processing. The researchers, led by Jin Gu of Tsinghua University together with co-first authors Xinqi Li and Yuhan Fan, represent a tissue as a graph in which cells are nodes connected to their spatial neighbors. Gene expression is then treated as a signal flowing across this graph. By applying a bank of heat kernel-based low-pass filters, the method smooths the expression signal over progressively larger diffusion distances, analogous to watching heat spread across a metal plate at different time points. Each filter setting corresponds to a distinct scale, from the immediate vicinity of a single cell up to sprawling tissue domains.

The technical elegance lies in how the method identifies which scales actually matter. Rather than requiring users to guess, SpatialZoomer constructs a cross-scale similarity map that quantifies how alike the clustering results are at every pair of scales. A dynamic programming procedure then partitions this map into blocks, and the boundaries between blocks mark the critical scales at which the tissue’s organization genuinely changes. This automatic scale selection is a departure from most prior multiscale approaches, which either demand manual parameter tuning or blend scales together without pinpointing where biologically meaningful transitions occur.

Once the critical scales are found, the software performs clustering at each one using a two-step strategy that combines K-Means pre-clustering with Leiden clustering. According to the authors, this hybrid approach captures the manifold structure of transcriptomic data while mitigating the distorting effects of spatial autocorrelation, the statistical tendency of neighboring cells to resemble one another simply because they are neighbors. The team demonstrated the method across nine datasets spanning more than 50,000 to over 1.1 million cells and multiple platforms, including cell-segmented and non-segmented Visium HD data and Stereo-seq data.

The biological payoffs are substantial. In an Alzheimer’s disease mouse brain dataset profiled with Xenium, SpatialZoomer detected a disease-progression-related signal that appeared only within a specific scale interval, around scale 6.0, and vanished at neighboring scales outside that interval. This scale-specific emergence suggests that some pathological signatures in neurodegeneration are properties of multicellular neighborhoods rather than of individual cells, a finding that single-cell-focused pipelines would likely miss or dilute. Sensitivity analyses confirmed that the signal was stable within its identified interval, reinforcing the reliability of the automatic scale detection.

Cancer biology provided equally striking demonstrations. In lung cancer data, the method identified macrophage subclusters whose identities depended on their spatial context, and in ovarian cancer it revealed spatially dependent subclusters of cancer-associated fibroblasts, the stromal cells that famously sculpt the tumor microenvironment. Perhaps most provocatively, at the niche scale the tool uncovered a trilayer organization in tumor cells, in which stressed, proliferative and epithelial-to-mesenchymal transition-like states are arranged in spatially structured layers. This stress-proliferation-EMT architecture echoes earlier observations of layered organization in glioblastoma and hints at a general principle of how aggressive tumors organize their own neighborhoods.

Benchmarking showed that SpatialZoomer achieves comparable or superior performance to existing single-scale methods on standard tasks such as spatial domain identification, while offering two practical advantages that could drive adoption. First, it is computationally frugal. Runtime and memory profiling across four devices, including a laptop with 32 gigabytes of RAM, showed that the method can handle datasets of roughly a million cells, with only the largest dataset exceeding laptop memory. The authors attribute this efficiency to the use of non-negative matrix factorization for dimensionality reduction and to efficient spectral filtering techniques, including Chebyshev polynomial approximations and Krylov subspace methods for computing matrix exponentials. Second, the method requires no specialized hardware, lowering the barrier for laboratories without access to GPU clusters.

The field has seen a flurry of multiscale and foundation-model approaches, including FICTURE for submicron-resolution analysis, BANKSY for unified cell typing and domain segmentation, CellCharter for niche discovery, and deep learning models such as Nicheformer and HEIST. SpatialZoomer distinguishes itself by making the scale itself an explicit, interpretable output rather than a hidden hyperparameter. The scale-to-distance mapping analyses in the paper show that the abstract scale parameter corresponds to effective physical diffusion distances that can be translated into micrometers, giving biologists a tangible sense of what each analytical zoom level means in tissue space.

The software, released as a Python package with tutorials at lifeome.net/software/spatialzoomer, integrates with the established single-cell ecosystem built around tools such as Scanpy and Squidpy. All processed data and reproduction code have been deposited on Zenodo, and the datasets analyzed, from mouse brain atlases to human lung and ovarian tumor specimens, are publicly available from their original sources. As spatial transcriptomics datasets balloon toward whole-organ scale, tools that can move fluidly between the cell, the niche and the domain, while telling the researcher exactly where the interesting scales lie, may become as indispensable as the microscopes and sequencers that generate the data. For now, SpatialZoomer offers a compelling demonstration that sometimes the most important discovery is not a new cell type, but the right vantage point from which to see what was there all along.

Subject of Research: A multi-scale computational method for analyzing spatial transcriptomics data at single-cell, niche and tissue-domain resolutions

Article Title: SpatialZoomer: multi-scale feature analysis of spatial transcriptomics

Article References: Li, X., Fan, Y., Han, Y., Guo, W., Huang, J., Yan, N., Chen, Z., Wu, Y., Miao, Y., Hou, L., Zhang, X., Chen, Z., & Gu, J. (2026). SpatialZoomer: multi-scale feature analysis of spatial transcriptomics. Nature Biomedical Engineering. https://doi.org/10.1038/s41551-026-01800-w

Image Credits: AI Generated

DOI: 10.1038/s41551-026-01800-w

Keywords: spatial transcriptomics, SpatialZoomer, graph signal processing, Alzheimer's disease, tumor microenvironment, single-cell analysis, cancer-associated fibroblasts, spatial domains, computational biology, bioinformatics, niche analysis, Nature Biomedical Engineering

The starting point for the research is the HAR model, one of the most influential tools in realized volatility forecasting. Realized volatility is estimated from intraday price data, typically by summing squared five-minute returns, and is far more informative than the squared daily returns used in older GARCH-style models. HAR summarizes the recent past with a handful of interpretable averages, typically over one, five, and twenty-two days, mirroring the daily, weekly, and monthly rhythms of market participants. Its simplicity makes it remarkably robust, but it has two structural blind spots. It is essentially univariate, ignoring how volatility transmits from one market to another, and its linear averages can miss nonlinear information hiding inside the look-back window, such as the particular shape of a volatility path that rises slowly and then collapses.

To handle cross-market transmission, the authors build on a 2025 framework called GSPHAR, which embeds a directed spillover network directly into the forecasting architecture. The network is constructed using the Diebold-Yilmaz connectedness framework: a vector autoregression with a lag order of 22 is fitted to standardized realized volatility, and a generalized forecast error variance decomposition with a 22-day horizon quantifies how much shocks in one market drive forecast uncertainty in another. Crucially, this graph is estimated only on the training sample to avoid look-ahead bias, then sparsified by keeping only the strongest quarter of links. Because the graph is directed, with volatility flowing asymmetrically between markets, the model uses the magnetic Laplacian, a spectral operator that encodes direction through complex phases, allowing a graph Fourier transform that preserves which way information travels.

The genuine innovation of Sig-GSPHAR lies in what it adds to this spectral backbone. For each market, the model takes the raw 22-day window of realized volatility, augments it with a normalized time channel to form a two-dimensional path, and computes a truncated path signature of depth four. Path signatures, rooted in the rough path theory developed by Terry Lyons and building on work by Kuo-Tsai Chen from the 1950s, convert a sequential path into a structured collection of iterated integrals. The first level captures simple increments, the second captures interactions between pairs of increments, and higher levels encode progressively richer temporal structure. A key theoretical property guarantees that linear functions of signatures can approximate any continuous nonlinear functional of the path, meaning the model can learn complex within-window dynamics without resorting to massive black-box recurrent networks.

The mathematics also dictates a careful trade-off. The magnitude of the depth-k signature terms scales with the k-th power of the path’s total variation, divided by k factorial. A single large jump inside the look-back window therefore inflates the higher-order coordinates, making them noisy and potentially misleading descriptions of what follows, since volatility typically mean-reverts after extreme spikes. This factorial bound explains why the authors cap the truncation depth at four: deeper signatures grow geometrically in dimension and inject noise rather than signal. An ablation study confirms the prediction, with depth four consistently delivering the best performance across all five volatility proxies, while depth two is too shallow to capture nonlinear interactions and depth five degrades generalization. A learnable sigmoid gate, shared across markets and time, further suppresses uninformative signature channels.

The empirical evaluation covers 29 stock market indices from June 2009 to May 2022, spanning 3,395 daily observations per market and five realized volatility proxies: five-minute realized variance, bipower variation, median realized volatility, realized kernel variance with a Parzen kernel, and realized semivolatility. All models were trained under an identical protocol, with the same chronological 70/30 split, the same 22-day look-back, and the same tuning budget, ensuring that performance differences reflect modeling choices rather than experimental design. For one-step-ahead forecasts, Sig-GSPHAR achieved the lowest error in 114 of 145 market-proxy pairs for mean squared error and 112 of 145 for mean absolute error. Paired one-sided Wilcoxon signed-rank tests with Holm correction for multiple comparisons confirmed that the improvements over GSPHAR are statistically significant for most proxies, with p-values as low as 0.006 for mean squared error on realized semivolatility.

The advantage becomes even more pronounced at the 22-day horizon, where the signature block, computed from the full look-back segment aligned with the long target, can encode nonlinear interactions that linear lag aggregation cannot. Sig-GSPHAR improved on GSPHAR in 26 of 29 markets for realized variance and in all 29 markets for four of the five proxies under mean absolute error. Holm-adjusted p-values fell to the order of 10 to the minus 6 for mean squared error and 10 to the minus 8 for mean absolute error, among the strongest statistical evidence reported in this literature. The result suggests that the longer the forecasting horizon, the more valuable it becomes to represent the shape of the volatility path rather than just its coarse averages.

Perhaps the most scientifically interesting part of the paper is its honest diagnosis of when signatures fail. The aggregate advantage of the new model comes almost entirely from calm days: on the 95 percent of test days with ordinary volatility levels, Sig-GSPHAR is more accurate, while on the top 5 percent of extreme days it slightly under-predicts true spikes, with a median forecast-to-realized ratio of 0.62 versus 0.68 for GSPHAR. The two markets where the new model loses to its predecessor, Finland’s OMXHPI and Denmark’s OMXC20, are precisely the most heavy-tailed in the panel, with sample kurtosis of 259 and 89 respectively. For OMXHPI, just ten days account for 43 percent of the entire out-of-sample squared error, so a small difference on a handful of high-leverage days determines the headline number. The authors trace this behavior to the graph-HAR backbone rather than the signature branch itself, and suggest jump-aware loss functions or variance-stabilized log domains as remedies.

Two robustness analyses strengthen the practical case. First, a rolling-window re-estimation of the spillover network shows that the dominant transmission structure is highly persistent: in normal periods the Spearman rank correlation between rolling and static net-spillover vectors is about 0.81, and even during the COVID-19 shock, when the correlation drops to roughly 0.56, total connectedness stays high and the set of top net-transmitters is largely preserved. Because the model uses the graph only through the magnetic-Laplacian spectral basis, which depends on the dominant structure rather than fragile individual edges, a static graph estimated once on the training data remains adequate. Second, the authors equip their point forecasts with distribution-free conformal prediction intervals using a volatility-normalized score. Both Sig-GSPHAR and GSPHAR achieve near-nominal coverage, but at matched coverage the new model produces intervals that are 7 to 11 percent narrower across all five proxies, translating its point-forecast gains into sharper, more informative uncertainty bands that widen during turbulence and tighten in calm markets.

The broader significance of the work lies in demonstrating that ideas from stochastic analysis can be fused productively with graph machine learning in a domain as noisy and adversarial as financial volatility. Rather than replacing interpretable econometrics with opaque deep networks, Sig-GSPHAR keeps the HAR skeleton, adds a mathematically principled nonlinear feature layer, and aligns everything in a single spectral representation governed by the physics of market spillovers. The authors point to time-varying spillover graphs and robust training losses as natural next steps, but the core message stands: when the goal is anticipating how turbulence propagates through a globally connected financial system, the shape of the recent past matters as much as its average, and rough path theory provides exactly the language needed to describe that shape.

Subject of Research: Multivariate realized volatility forecasting using path signatures and directed graph spectral methods

Article Title: Signature-enhanced graph spectral HAR for multivariate realized volatility forecasting

Article References: Guo, Z., Han, A., Wang, C., & Gao, J. (2026). Signature-enhanced graph spectral HAR for multivariate realized volatility forecasting. Applied Intelligence, 56(14), Article 406. https://doi.org/10.1007/s10489-026-07445-1

Image Credits: AI Generated

DOI: 10.1007/s10489-026-07445-1

Keywords: realized volatility, path signatures, rough path theory, graph signal processing, HAR model, magnetic Laplacian, Diebold-Yilmaz spillover, volatility forecasting, conformal prediction, financial econometrics, machine learning, spillover networks