GeneFlow: Translation of Single-cell Gene Expression to Histopathological Images via Rectified Flow

TL;DR

GeneFlow is a novel framework that translates single-cell gene expression data into high-resolution histopathological images using rectified flow, enabling insights into cellular structures and disease patterns.

Contribution

It introduces a rectified flow-based approach for mapping transcriptomics to images, improving over diffusion-based methods and allowing realistic cellular morphology generation.

Findings

Outperforms diffusion-based baseline in all experiments

Generates realistic cellular morphology features

Reveals dysregulated patterns for disease diagnosis

Abstract

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.