Organizations: Vanderbilt University, Nashville TN 37235, USA · Weill Cornell Medicine, New York, NY 10065, USA · Vanderbilt University Medical Center, Nashville TN 37232, USA · The University of Texas MD Anderson Cancer Center, Houston TX 77030, USA
Spatial transcriptomics (ST) platforms simultaneously measure gene expression and capture cell nucleus images at single-cell resolution. A fundamental property of tissue preparation introduces a confounding relationship between the two measurement modalities: sectioning a three-dimensional tissue into a thin slice produces a two-dimensional cross-section whose area depends on the cutting angle. Larger cross-sections expose more cytoplasm, yielding higher mRNA counts, while pathomics features such as nuclear area are measured directly from the same 2D boundary. This sectioning geometry creates an apparent correlation between gene expression and pathomics that reflects the measurement process rather than independent biology. Existing gene-to-image methods generate cell images from gene features but do not support condition-driven editing of existing images across both omics layers. We apply the Manifold-Aware Diffusion (MAD) framework to this dual-omics editing setting, enabling targeted modification of one feature group while preserving the other. Evaluated on podocyte cells from CosMx SMI kidney tissue, MAD achieves MAE=0.6636 and FID=25.19 for pathomics editing, outperforming all compared baselines. A controlled simulation study further confirms that the VAE component in MAD better preserves cross-group independence than CCDM (MAD without VAE).
Figures & tables
Figure 1: (a) Tissue sectioning mechanism: a 3D cell is cut at varying angles, producing 2D cross-sections of different areas. Larger cross-sections yield higher mRNA counts and larger nuclear area measurements, creating a sectioning-induced correlation between gene expression and pathomics. (b) Existing gene-to-image models are generation-only and condition on a single omics modality; no prior method performs condition-driven editing jointly across gene expression and pathomics. (c) Our approach applies MAD [ 13 ] to dual-omics conditional editing: given an existing cell nucleus image, one feature group is edited to a target value while the other is preserved.
Figure 2: Overview of the dual-omics conditional image editing framework. Gene expression g and pathomics p are concatenated into a joint feature vector c . The MAD framework [ 13 ] trains a conditional diffusion model on c and a feature manifold VAE that encodes within-group correlations. At editing time, a target vector c∗ is projected through the VAE to obtain a manifold-consistent conditioning signal, which is passed to the diffusion model to generate the edited cell nucleus image.
Figure 3: Pairwise Pearson correlation matrix for the 8 features (3 genes + 5 pathomics) in the podocyte dataset. Strong within-group correlations are observed (Area–Perimeter r=0.95 ; IntensityStd–IntensityMean r=0.76 ), while cross-group correlations remain near zero ( ∣r∣≤0.14 ).
Figure 4: Simulation study: cross-group independence preservation. Δ - Δ scatter plots show the editing-induced change in Group 1 features (Area, Gene1). Within-group pairs (diagonal blocks) show high correlation; cross-group pairs (off-diagonal blocks) should approach zero for a method that preserves independence. MAD achieves better cross-group independence preservation (cross-group MAE = 0.112 when editing Area) compared to CCDM without VAE (cross-group MAE = 0.271).
Figure 5: Pathomics editing: generation baselines versus MAD. Each row shows a sweep from low (left) to high (right) target values along the Area axis (top) and Intensity axis (bottom) for GeneFlow, Genes2Morphology, and MAD. Generation baselines produce outputs that vary with gene input but do not follow the specified pathomics gradient. MAD produces progressive morphological change along both editing dimensions.
Figure 6: Pathomics editing: editing baselines versus MAD. Stable Diffusion (SD) produces minimal visible change across editing steps. CCDM (MAD without VAE) shows abrupt intensity variations without smooth structural progression. MAD achieves smooth, controlled morphological changes along both the Area and Intensity editing axes.
Type
Method
MAE ↓
FID ↓
SSIM ↑
Generation
GeneFlow [ 11 ]
0.8807
55.47
0.5756
Genes2Morphology [ 7 ]
1.659
289.26
0.1027
Editing
Stable Diffusion [ 9 ]
1.901
63.21
0.8415
CCDM [ 1 ]
0.9158
31.15
0.8078
MAD [ 13 ]
0.6636
25.19
0.7964
Table 1: Quantitative pathomics editing results. MAE measures editing accuracy (target vs. measured pathomics values of generated images); FID measures generated image quality; SSIM measures structural fidelity to the original cell image. ↓ lower is better; ↑ higher is better.
Figure 7: Gene editing results: ACTN4 = 0 (normal, left) vs. ACTN4 = 5 (disease-associated, right) for Genes2Morphology, GeneFlow, Stable Diffusion, and MAD. Generation models (Genes2Morphology, GeneFlow) produce outputs without a reference image. Editing models (SD, MAD) modify an existing cell image. MAD produces visually distinct outputs between the two ACTN4 conditions.
Predicting spatial gene expression from hematoxylin and eosin (H&E)-stained images offers a cost-effective alternative to spatial transcriptomics (ST). However, existing methods treat H&E images as generic visual inputs and ignore their intrinsic biological hierarchy, where spatially organized cell types collectively form functional tissue microenvironments that govern local gene expression programs. To bridge this gap, we formulate H&E-to-ST prediction as a cross-modal semantic translation task and propose Path2ST, a hierarchically grounded autoregressive framework featuring three key components: (i) a Hierarchical Cell-Tissue Conditioning mechanism that fuses explicit and implicit cellular features with tissue-level semantic representations to construct hierarchical conditioning signals; (ii) a Scale-Adaptive Autoregressive Generation process over a hierarchical semantic vocabulary, enabling coarse-to-fine, biologically consistent expression synthesis; and (iii) SpectraLoss, a full-spectrum objective that jointly enforces ordinal fidelity, models transcriptional bursts, and aligns semantic structures with cell types. Extensive experiments on three datasets demonstrate state-of-the-art performance, validating that Path2ST generates highly accurate and spatially coherent transcriptomic profiles. The related code is released at https://github.com/RuochenLiu23/Path2ST.
Ruochen Liu, Wei Lou
Faculty of Science and Engineering University of Liverpool Liverpool, United Kingdom · College of Mathematical Medicine Zhejiang Normal University Jinhua, China
Spatial transcriptomics measures gene expression while preserving where each cell sits in the tissue. The inverse, what a given transcriptomic state looks like, is less explored but equally useful: it lets researchers visualize the morphological consequence of a molecular state and interpret a spatial assay's molecular calls against real tissue structure. GeneFlow, the current state of the art on the Xenium gene expression-to-image benchmark, trains a model from scratch, incurring a high cost on every new target dataset it is applied to. We instead propose GESTALT, which reformulates gene expression-to-image synthesis as predicting a pathology foundation model's image embedding. A lightweight gene-token transformer predicts this embedding from multi-cell gene expression, and a frozen or LoRA-adapted diffusion decoder (PixCell) renders it as an image. Because this visual knowledge is reused rather than relearned, GESTALT trains in a fraction of the time. In both multi-cell and single-cell settings, our method outperforms the SoTA on perceptual metrics across all three Xenium datasets. Beyond improved image quality, GESTALT natively synthesizes at 1024 x 1024, and is evaluated in three generalization settings: zero-shot transfer across gene panels and datasets, held-out patients and disease states, and transfer across studies and tissue sources. Coupled with a statistical model of spatial expression, GESTALT also provides, to our knowledge, the first paired synthesis of cell-resolved transcriptomic fields and tissue images with explicit control over cell type, density, and composition. Code is available at https://github.com/trongnhandev1011/Gestalt.
Spatial transcriptomics (ST) enables genome-wide gene expression profiling while preserving tissue architecture, but its cost and limited scalability remain major bottlenecks. This has motivated models that predict spatial expression directly from routine histology. Despite promising results, most existing approaches operate at the gene level without leveraging established transcriptomic modeling practices and rely on heterogeneous gene selection strategies, which complicates fair comparison across methods. We propose to reformulate morphology-to-transcriptomics prediction as conditional generation in transcriptional program space, thereby exploiting coordinated transcriptional variation instead of predicting genes independently. Using consensus non-negative matrix factorization (cNMF), we extract a low-dimensional set of transcriptional programs capturing coordinated expression variation in the training data, and train a conditional diffusion model to generate program activations from histology. This formulation exploits coordinated transcriptional variation and substantially lowers the dimensionality of the conditional generative task.
Ruyter Swann, Dorent Reuben, Racoceanu Daniel
Sorbonne Université, CNRS, Inserm, AP-HP, Inria, Paris Brain Institute – ICM, Paris, France