TBNAI
From Cross-Section to Cross-Omics: Conditional Cell Nucleus Image Editing in Spatial Transcriptomics
Publication: cs.AI updates on arXiv.org
Author: Yuechen Yang, Junchao Zhu, Junlin Guo, Yanfan Zhu, Ruining Deng, Zhengyi Lu, Tianyuan Yao, Chongyu Qu, Marilyn Lionts, Xingyi Guo, Yaohong Wang, Yu Wang, Shilin Zhao, Haichun Yang, Yuankai Huo
Source published: 03/11/2026, 05:00:00 UTC
Briefing created: 03/11/2026, 10:45:46 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 03/11/2026, 09:03:25 UTC.
Briefings
A new method called Manifold-Aware Diffusion (MAD) edits spatial transcriptomics and cell nucleus images simultaneously using conditional diffusion. Evaluated on kidney podocytes from CosMx SMI tissue, MAD reduced Mean Absolute Error (MAE) to 0.6636 and FID to 25.19 for pathomics editing, significantly outperforming existing baselines. A simulation study confirmed that the VAE component in MAD preserves independence between gene expression and nuclear measurements better than alternative approaches.
Key points
- The Manifold-Aware Diffusion (MAD) framework edits both spatial transcriptomics and cell nucleus images simultaneously using conditional diffusion.
- MAD achieved MAE=0.6636 and FID=25.19 for pathomics editing on kidney podocytes from CosMx SMI tissue, outperforming all compared baselines.
Context
Spatial transcriptomics platforms measure gene expression and capture cell nucleus images at single-cell resolution, but tissue sectioning geometry creates a confounding relationship between the two modalities. Existing methods do not support condition-driven editing across both omics layers, and MAD addresses this by enabling targeted modification of one feature group while preserving the other.
Limitations
The evaluation is limited to podocyte cells from CosMx SMI kidney tissue. A controlled simulation study confirmed the VAE component in MAD better preserves cross-group independence than CCDM, but the specific biological impact of this improvement on other tissue types remains unverified.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 03/11/2026, 10:45:46 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeihbohr42w2jes6n7aszkv4ukjgtipvkk57zvuqtilfkh7dk3f6mba