Geometry-First Generative Spatial Single-Cell Reconstruction

2026-05-27

AI Summary

This study addresses the limitations of single-cell RNA sequencing, which lacks spatial context, and spatial transcriptomics, which suffers from low resolution, as well as the poor generalizability of existing integration methods that rely on fixed grids and sample-specific coordinate systems in the absence of paired data. To overcome these challenges, the authors propose GEARS, a novel framework that aligns single-cell and spatial data through a domain-invariant encoder and reconstructs intrinsic cellular spatial geometry using a permutation-equivariant generator coupled with a diffusion refinement module—without requiring cell-type labels or tissue images. GEARS introduces a “geometry-first” paradigm that eliminates dependence on fixed grids, explicit cell-to-spot assignments, and auxiliary modalities, enabling pose-invariant, cross-section generalizable, high-fidelity spatial reconstruction. It significantly outperforms current baselines in preserving global distances, local neighborhood structures, and overall spatial alignment.

Abstract

Single-cell RNA sequencing (scRNA-seq) profiles large numbers of cells but loses spatial context, whereas spatial transcriptomics (ST) preserves partial spatial structure at lower resolution. Most existing integration methods either deconvolve spot mixtures or map cells onto a measured spot lattice, which ties reconstructions to a fixed grid and slide-specific coordinate systems, a limitation that is especially problematic in unpaired settings. We propose GEARS, a geometry-first framework that reconstructs an intrinsic single-cell spatial geometry guided by ST, without relying on cell-type labels, histological images, or cell-to-spot assignment. GEARS first learns a domain-invariant expression encoder that aligns ST spots and dissociated cells, and then trains a permutation-equivariant generator with a diffusion-based refiner with EDM-style preconditioning to generate local spatial geometries under pose-invariant supervision derived from ST coordinates. At inference, GEARS reconstructs geometry on many overlapping subsets of scRNA-seq cells, aggregates predicted pairwise distances across subsets, and solves a global distance-geometry problem to obtain canonical two-dimensional coordinates and a dense distance matrix. Extensive quantitative and qualitative experiments, including cross-section generalization, show that GEARS consistently improves global distance preservation, local neighborhood fidelity, and spatial distribution alignment compared to strong spatial mapping and deconvolution baselines.

Problem

• spatial transcriptomics
• single-cell RNA sequencing
• spatial reconstruction
• geometry inference
• unpaired integration

Innovation

• geometry-first
• spatial single-cell reconstruction
• domain-invariant encoding
• permutation-equivariant generation
• distance geometry