PPubMed29 Jun 2026

ARISE: RNA-anchored shared-edge topology and hierarchical fusion for spatial multi-omics integration

Xiangxiang Wang, Yanchi Su, Gaoyang Hao, Meng Wang, Yunhe Wang, Xiangtao Li, Anthony Mathelier

Abstract

As illustrated in [btag465-F1], ARISE begins with modality-specific preprocessing to harmonize heterogeneous spatial multi-omics inputs before graph construction. RNA data undergo gene- and cell-level quality control, followed by selection of the top 3000 highly variable genes, total-count normalization, log1p transformation, and centering. ADT measurements are normalized independently for each spot using centered log-ratio normalization and then centered, whereas ATAC profiles are processed with TF-IDF normalization followed by latent semantic indexing. The resulting feature matrices are then

A figure from ARISE: RNA-anchored shared-edge topology and hierarchical fusion for spatial multi-omics integration — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

ARISE uses an RNA‑anchored shared‑edge graph to integrate spatial multi‑omics, achieving more accurate and interpretable spatial domain identification across bi‑ and tri‑modal datasets.

What’s new

i.Introduces a shared‑edge topology by intersecting RNA feature‑similarity and spatial‑proximity graphs
ii.Develops an inside‑out hierarchical fusion module to combine RNA and auxiliary modalities
iii.Provides theoretical proof that graph intersection minimizes false‑positive edges in k‑of‑r fusion rules
iv.Demonstrates superior clustering and spatial domain recovery on bi‑modal and tri‑modal benchmarks

How they did it

ARISE first constructs separate RNA similarity and spatial graphs, intersecting them to form a high‑confidence scaffold. Auxiliary modalities are encoded on this shared graph via GCNs, while RNA is encoded on both graphs. The resulting embeddings are fused hierarchically inside‑out and trained with spatial regularization, reconstruction losses, and L1/L2 penalties.

What they found

Human Lymph Node: AMI = 0.4141, NMI = 0.4182, ARI = 0.3427 (best among compared methods)
Mouse Thymus (RNA+ADT): Silhouette Coefficient = 0.2680 (strong spatial coherence)
Peak‑gene correlation median: 0.38 in differentiated cells vs 0.32 in neural progenitors (demonstrating regulatory coupling)

Caveats

✕Relies on RNA as the sole anchor modality, limiting applicability when RNA quality is poor
✕Conservative shared‑edge design may discard useful edges present in only one modality
✕Currently designed for single‑sample spatial graphs, not yet extended to cross‑sample or image‑informed settings

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Similar papers in all four archives

The same idea searched in arXiv, PubMed, bioRxiv and chemRxiv at once, so a method can lead you to where it is being applied.

AarXiv4 found

PPubMed2 found

BbioRxiv4 found

CchemRxiv2 found

Searching the other archives…