ARISE

ARISE: RNA-Anchored Shared-Edge Topology and Hierarchical Fusion for Scalable Spatial Multi-Omics Integration

Anchored RNA for Integrated Spatial Embedding

---

Overview

ARISE is an RNA-anchored framework for spatial multi-omics integration. Rather than constructing independent modality-specific graphs, ARISE defines a shared-edge topology by intersecting an RNA feature-similarity graph with a spatial-proximity graph, retaining only edges supported by both transcriptional similarity and physical adjacency. Auxiliary modalities (ADT, ATAC, histone modifications) are encoded on this common scaffold, and an inside-out hierarchical fusion module integrates them into a unified latent representation.

Key advantages:

• Theoretically grounded: graph intersection minimizes false-positive edges across all k-of-r fusion rules (Theorems 1–3)
• Stable under perturbation: GNN encoder drift is provably linear in graph perturbation magnitude (Theorem 4)
• Modular: additional modalities can be incorporated without redefining the shared-edge topology
• Supports bi-modal (RNA+ADT, RNA+ATAC) and tri-modal (RNA+ATAC+Protein/Histone) settings

---

Installation

Requirements

• Python >= 3.8
• PyTorch >= 1.12
• CUDA (recommended)

Dependencies

torch>=1.12.0
torch-geometric
scanpy>=1.9.0
anndata>=0.8.0
numpy
scipy
pandas
scikit-learn
matplotlib
seaborn

Setup

git clone https://github.com/XiangxiangWang-code/ARISE.git
cd ARISE
pip install -r requirements.txt

---

Data Preparation

ARISE accepts input in AnnData .h5ad format. Each modality should be stored as a separate AnnData object with spatial coordinates in adata.obsm['spatial'].

Data Format

import anndata as ad

# RNA modality
adata_rna = ad.read_h5ad("rna.h5ad")         # shape: (n_spots, n_genes)

# Auxiliary modality (ADT or ATAC)
adata_adt = ad.read_h5ad("adt.h5ad")         # shape: (n_spots, n_proteins)

# Spatial coordinates must be stored in obsm
# adata_rna.obsm['spatial'] — shape: (n_spots, 2)

---

Reproducing Paper Results

Training scripts are located in ARISE/code/.

RNA + ADT: Human Lymph Node

python ARISE/code/HLN.py

---

Contact

For questions or issues, please open a GitHub Issue or contact:

• Xiangxiang Wang: 2319659260@qq.com