Spatial transcriptomics (ST) provides spatially resolved gene expression measurements, but most 3D ST datasets are acquired as independently processed 2D sections. Computational approaches span slice registration and integration, explicit or generative interpolation, and coordinate-field modeling. Within the last family, implicit neural representations (INRs) offer a continuous, coordinate-queryable parameterization, but standard isotropic coordinate encodings do not explicitly reflect that serial-section measurements are dense in-plane and sparse along the axial direction. We present HoloGene, an anisotropic conditional INR framework for estimating continuous 3D gene-expression fields from pre-registered ST sections. Its central contribution is an acquisition-aware factorization that couples dense in-plane synthesis with sparse axial evolution using established GAE, FiLM, pseudo-label supervision, and CORAL components. HoloGene compresses high-dimensional expression profiles into a graph-autoencoder latent manifold, uses pseudo-label-derived signature conditioning to provide coarse intra-section context, and applies depth-dependent FiLM modulation to model axial variation. A signature-stratified correlation-alignment regularizer encourages cross-slice covariance alignment in latent features. Using held-out measured sections as cross-sectional references, HoloGene improves reconstruction accuracy and biological preservation metrics compared with general-purpose INRs and adapted spatial-transcriptomics baselines. These results support anisotropic conditional modeling as a practical strategy for estimating 3D expression fields from discontinuous ST sections. Code is available at https://github.com/Szym29/HoloGene.
Figures & tables
Figure 1: HoloGene reconstructs holistic 3D gene fields from discontinuous ST sections. By lifting sparsely sampled 2D slices into a continuous 3D field, HoloGene enables volumetric downstream analysis, including spatial-domain recovery, cell–cell communication inference, and pathway enrichment.
Figure 2: Overview of HoloGene. We first compress high-dimensional gene expression into latent embeddings via a graph autoencoder, then use PSN to distill training-expression partitions into coarse signature guidance, and synthesize a continuous 3D latent field using a decoupled, FiLM-conditioned INR with CORAL regularization to handle inter-slice anisotropy and regularize slice-specific representation variation.
Figure 3: Ablation on LCORAL : UMAP visualizations colored by section ID and anatomical regions on the whole mouse brain data ( Zhang et al., 2023 ) . Enabling LCORAL leads to better cross-slice mixing and less visual separation by section ID.
Figure 4: Held-out section synthesis and inter-section querying. A) Marker-gene and domain-map reconstruction on held-out measured sections. Domain maps show ground-truth annotations in the first row and predicted-expression clusters mapped to annotation colors in the remaining rows. See full results in Appendix A . B) Querying HoloGene between measured sections provides model-based interpolation estimates with smooth STMN2 variation. See full results in Appendix C .
Figure 5: HoloGene preserves CCC patterns in MOSTA E16.5. (A) Ground-truth and HoloGene-predicted communication scores with their differences in the testing set. (B) Overlap rate of top significant ligand-receptor interactions recovered by HoloGene across annotation pairs. Darker blue indicates higher recovery.
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 6: Qualitative results of virtual slice synthesis on representative 3D samples. We visualize measured reference slices and the corresponding predictions from different methods; marker-level outcomes vary across genes and sections, while aggregate comparisons are reported in the main text.
Figure 7: Leiden-signature maps obtained from predicted gene expressions. HoloGene yields more spatially coherent tissue regions with fewer isolated mislabeled spots compared with baselines.
Figure 8: Annotation-referenced CCC network preservation on (A) MOSTA E16.5 and (B) whole mouse brain. Nodes denote anatomical annotations, and directed edges show top-scoring sender-to-receiver ligand–receptor communication connections. Ground-truth and HoloGene-predicted networks show similar high-confidence communication topology, indicating that virtual-slice synthesis preserves higher-order interaction structure.
Figure 9: Visualization of virtual-slice evolution for two gene markers. The leftmost and rightmost panels show measured slices, and the intermediate panels show synthesized virtual slices.
Figure 10: Annotation-referenced cell–cell communication (CCC) preservation on the two primary datasets. Rows and columns denote anatomical annotations used for post hoc sender–receiver grouping. For each dataset, we compare ground-truth CCC interaction scores, HoloGene-predicted CCC interaction scores, prediction-minus-ground-truth residuals, and top ligand–receptor interaction overlap rates. (A) MOSTA E16.5. (B) Whole mouse brain. HoloGene recovers similar label-pair communication structures to the ground truth, with small residual differences and high overlap of top ligand–receptor interactions across most annotation pairs.
Three-dimensional spatial transcriptomics (3D ST) enables the analysis of gene-expression patterns across tissue depths. However, 3D ST measurements are acquired from discrete 2D tissue sections, which are often sparsely sampled due to high acquisition costs. To address this limitation, we propose a histology-informed multi-task implicit neural representation (HIM-INR) for continuous 3D gene-expression reconstruction from sparse ST observations. HIM-INR jointly reconstructs gene-expression profiles and histological representations using an INR with a shared backbone, leveraging densely sampled histological images as auxiliary spatial supervision. Once trained, HIM-INR predicts gene expression at arbitrary 3D coordinates without requiring histological images. Experiments on two 3D ST datasets demonstrate improved reconstruction over existing methods, particularly as the number of missing ST sections increases.
A 3D transcriptomic profile of a tissue can give a holistic view of its organization in health and disease, but spatial transcriptomics (ST) is costly, so usually only some serial sections are profiled, while H&E is available for all. Most methods that infer ST from histology treat each section independently and ignore 3D structure, while existing 3D-aware approaches are not generative and do not scale well. We present Holographic Tissue Expression Inpainting and Analysis (HoloTea), a 3D-aware flow-matching framework that imputes spot-level gene expression on an unprofiled section from its H&E and the measured expression of adjacent profiled sections. Our key idea is to take the measured expression at the same registered position on neighboring sections and feed this cross-section context to the flow model through a lightweight ControlNet, so that predictions follow anatomical continuity along z. To capture the discrete, zero-inflated nature of counts, the flow starts from a learned zero-inflated negative binomial (ZINB) prior that, on an unprofiled section, is predicted from the measured expression of the nearest profiled section. A global attention block with a small set of inducing tokens adds whole-slide context at a cost linear in the number of spots, so training and inference scale to large 3D ST datasets. Across three ST datasets spanning different tissue types and resolutions, HoloTea consistently improves 3D expression accuracy and generalization compared to 2D and 3D baselines. We envision HoloTea advancing the creation of accurate 3D virtual tissues, ultimately accelerating biomarker discovery and deepening our understanding of disease. Code is available at https://github.com/Lotfollahi-lab/HoloTea.
Mohammad Vali Sanian, Arshia Hemmat, Amirhossein Vahidi +10
Computer Science Department, University of Helsinki · Institute for Molecular Medicine Finland, University of Helsinki · Wellcome Sanger Institute +4
Spatial transcriptomics (ST) measures mRNA expression while preserving spatial organization, but multi-slice analysis faces two coupled difficulties: large non-rigid deformations across slices and inter-slice batch effects when alignment and integration are treated independently. We present INST-Align, an unsupervised pairwise framework that couples a coordinate-based deformation network with a shared Canonical Expression Field, an implicit neural representation mapping spatial coordinates to expression embeddings, for joint alignment and reconstruction. A two-phase training strategy first establishes a stable canonical embedding space and then jointly optimizes deformation and spatial-feature matching, enabling mutually constrained alignment and representation learning. Cross-slice parameter sharing of the canonical field regularizes ambiguous correspondences and absorbs batch variation. Across nine datasets, INST-Align achieves state-of-the-art mean OT Accuracy (0.702), NN Accuracy (0.719), and Chamfer distance, with Chamfer reductions of up to 94.9% on large-deformation sections relative to the strongest baseline. The framework also yields biologically meaningful spatial embeddings and coherent 3D tissue reconstruction. The code will be released after review phase.
Bonian Han, Cong Qi, Przemyslaw Musialski +1
New Jersey Institute of Technology, Newark, NJ, USA