Go to ICLR 2026 Conference homepageSPATIA: Multimodal Model for Prediction and Generation of Spatial Cell Phenotypes
Keywords: Spatial transcriptomics, Multimodal Model, Spatially conditioned morphology generation, Contrastive flow matching
TL;DR: We introduce SPATIA, a multi-scale generative and predictive model for spatial transcriptomics that integrates cell morphology, gene expression, and spatial context, evaluating on diverse biological tasks.
Abstract: Understanding how cellular morphology, gene expression, and spatial organization jointly shape tissue function is a central challenge in biology. Image-based spatial transcriptomics technologies now provide high-resolution measurements of cell images and gene expression profiles, but existing methods typically analyze these modalities in isolation or at limited resolution.
We address the problem by introducing SPATIA, a multi-scale generative and predictive model that learns unified, spatially aware representations by fusing morphology, gene expression, and spatial context from single-cell to tissue level. SPATIA incorporates a spatially conditioned image-to-image generation module that predicts cell morphologies under perturbations, enabling the study of microenvironment-dependent morphological changes such as tumor progression, immune remodeling, and subtype transitions.
We assembled a multi-scale dataset consisting of $17$ million cell-gene pairs, $1$ million niche-gene pairs, and $10,000$ tissue-gene pairs across diverse tissues and disease states. We benchmark SPATIA against $16$ existing models across $12$ individual tasks, which span several categories including cell annotation, cell clustering, gene imputation, cross-modal prediction, and image generation. SPATIA achieves improved performance over baselines and generates realistic cell morphologies that reflect transcriptomic perturbations.
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Submission Number: 3496
Loading