GeneFlow: Translation of Single-cell Gene Expression to Histopathological Images via Rectified Flow
This work addresses the challenge of aligning transcriptomics with cellular morphology in spatial transcriptomics, enabling realistic image generation and potential applications in disease diagnosis, though it is incremental as it builds on existing flow-based methods.
The authors tackled the problem of mapping single-cell gene expression to histopathological images by developing GeneFlow, a framework that uses rectified flow to generate high-resolution images with different staining methods, outperforming diffusion-based baselines in all experiments.
Spatial transcriptomics (ST) technologies can be used to align transcriptomes with histopathological morphology, presenting exciting new opportunities for biomolecular discovery. Using ST data, we construct a novel framework, GeneFlow, to map transcriptomics onto paired cellular images. By combining an attention-based RNA encoder with a conditional UNet guided by rectified flow, we generate high-resolution images with different staining methods (e.g. H&E, DAPI) to highlight various cellular/tissue structures. Rectified flow with high-order ODE solvers creates a continuous, bijective mapping between transcriptomics and image manifolds, addressing the many-to-one relationship inherent in this problem. Our method enables the generation of realistic cellular morphology features and spatially resolved intercellular interactions from observational gene expression profiles, provides potential to incorporate genetic/chemical perturbations, and enables disease diagnosis by revealing dysregulated patterns in imaging phenotypes. Our rectified flow-based method outperforms diffusion-based baseline method in all experiments. Code can be found at https://github.com/wangmengbo/GeneFlow.