Research map / Image generation / Autoregressive and latent generation
SPATIA: Multimodal Generation and Prediction of Spatial Cell Phenotypes
Zhenglun Kong, Mufan Qiu, John Boesen, xiang lin, Sukwon Yun, Tianlong Chen, Manolis Kellis, Marinka Zitnik
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add confidence-aware OT reweighting to flow matching (ablation) | SPATIA base · Conditional morphology generation | FID: 59.5 -> 59.1 | Improved |
| add morphology-profile alignment loss (ablation) | SPATIA + Reweight · Conditional morphology generation | Wasserstein correlation: 0.91 -> 0.94 | Improved |
| add morphology-profile alignment loss (ablation) | SPATIA + Reweight · Conditional morphology generation | KS: 0.62 -> 0.65 | Improved |
| add reconstruction (MAE) loss (ablation) | cell-level backbone · Cell classification | accuracy: 0.93 -> 0.94 | Improved |
| add multi-level (niche and tissue) hierarchy (ablation) | cell-level model with MAE loss · Cell classification | accuracy: 0.94 -> 0.97 | Improved |
| add cross-attention fusion (ablation) | multi-level SPATIA · Cell classification | accuracy: 0.97 -> 0.98 | Improved |
| add niche-level context (ablation) | SPATIA generation · Conditional morphology generation | FID: 60.3 -> 59.2 | Improved |
| corrupt 10-20% of OT pairs randomly (ablation) | SPATIA flow module · Conditional morphology generation | FID: 59.2 -> 61.0 -> 63.8 | Hurt |
| use SPATIA instead of single-cell baselines | cell clustering embeddings · CosMx | NMI: 0.560 -> 0.490 | Mixed |
| use SPATIA embeddings instead of UNI | BCNB receptor status prediction · BCNB ER AUC | AUC: 0.891 -> 0.902 | Improved |
| use SPATIA instead of CellPLM | cell annotation · MS dataset | F1: 0.709 -> 0.725 | Improved |
Closest papers by meaning
- Diffusion Generative Modeling for Spatially Resolved Gene Expression Inference from Histology Images (ICLR 2025 · similarity 0.89)
- Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching (ICML 2025 · similarity 0.88)
- Modeling Microenvironment Trajectories on Spatial Transcriptomics with NicheFlow (NeurIPS 2025 · similarity 0.88)
- SToFM: a Multi-scale Foundation Model for Spatial Transcriptomics (ICML 2025 · similarity 0.87)
- Preserving DEG Rankings for Gene Discovery in Histology-Based Spatial Gene Expression Prediction (NeurIPS 2026 · similarity 0.87)
- HEST-1k: A Dataset For Spatial Transcriptomics and Histology Image Analysis (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
