LATTICE：面向多模态空间组学整合的图自监督学习
英文标题：LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration
空间组学研究正日益融合转录组与表观组数据，但下游分析仍以单模态流程为主。本文提出的 LATTICE 框架采用图自监督学习，在单个点上对齐五种模态（Visium RNA、scMultiome RNA/ATAC、空间 ATAC 与空间 CUT&Tag），并利用 TransformerConv 编码器进行掩码重建与跨模态对齐训练。在一个包含 54,912 个点的私有黑色素瘤队列上，加入多模态后，与 Space Ranger 聚类的一致性（ARI 提升 0.157、NMI 提升 0.143）和空间连续性（提升 0.174）均显著改善；额外模态进一步提升了空间连续性与多模态效用分数，但可能降低与 RNA 参考标签的一致性，表明嵌入捕捉了超越转录组相似性的染色质与调控结构。该工作为多模态空间组学整合提供了实证可行的范式，同时也揭示了引入更强监督与外部基准的必要性，对多模态生物信息学工具开发者具有参考价值。
2.1 KB
612 B
380 B
9.1 KB
32.7 KB
22.8 KB
24.2 KB
34.7 KB
中文
[提交于 2026年7月15日]
Title: LATTICE：面向多模态空间组学整合的图自监督学习
查看论文《LATTICE：面向多模态空间组学整合的图自监督学习》的 PDF，作者 Jagan Mohan Reddy Dwarampudi 等 4 位作者
摘要：空间分辨组学研究日益结合转录组和表观基因组分析，但下游分析仍常使用单模态流程。我们提出了 LATTICE（Latent Alignment of Tissue-level and Transcriptomic Information for Cross-modal Embedding，基于组织级与转录组信息的跨模态对齐嵌入），一种图自监督框架，可从整合的多模态特征中学习点级别的表示。LATTICE 为每个 Visium 点集成了五个对齐的模态模块：Visium RNA、scMultiome RNA、scMultiome ATAC、空间 ATAC 和空间 CUT&Tag。这些模态在一个统一的格点表示中捕获了空间转录组测量、单细胞推断的调控活性以及原位染色质和组蛋白状态。LATTICE 构建了一个空间邻域图，并使用掩码重建、跨模态对齐和空间平滑目标来训练 TransformerConv 编码器。在一个来自匿名临床合作者的私有 11 样本黑色素瘤队列中（共计 54,912 个点），LATTICE 展示了稳定的优化行为、跨分析种子的可重现嵌入以及所有样本的完整多模态整合。仅在 Visium RNA 的基础上添加 scMultiome RNA 便显著提高了 11 次运行中与 Space Ranger 聚类的一致性（调整兰德指数 [ARI] +0.157，归一化互信息 [NMI] +0.143，空间连续性 +0.174）。额外模态进一步改善了空间连续性和多模态效用分数（MUS），尽管有时会降低与 RNA 衍生参考标签的一致性，这可能是因为学习到的嵌入捕捉到了超越单纯转录组相似性的染色质和调控结构。这些结果将 LATTICE 定位为一个实用且基于实证的多模态空间组学整合框架，同时也突显了对更强监督和更广泛外部基准的需求。
提交历史
来自：Jagan Mohan Reddy Dwarampudi [ 查看邮件 ]
[v1] 周三，2026年7月15日 22:54:47 UTC (3,779 KB)
English / 原文
[Submitted on 15 Jul 2026]
Title:LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration
View a PDF of the paper titled LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration, by Jagan Mohan Reddy Dwarampudi and 4 other authors
Abstract:Spatially resolved omics studies increasingly combine transcriptomic and epigenomic assays, yet downstream analysis is often still performed using single-modality pipelines. We present LATTICE (Latent Alignment of Tissue-level and Transcriptomic Information for Cross-modal Embedding), a graph-based self-supervised framework that learns spot-level representations from harmonized multimodal features. LATTICE integrates five aligned modality blocks per Visium spot: Visium RNA, scMultiome RNA, scMultiome ATAC, spatial ATAC, and spatial CUT&Tag. These modalities capture spatial transcriptomic measurements, single-cell inferred regulatory activity, and in situ chromatin and histone states within a unified lattice representation. LATTICE constructs a spatial neighborhood graph and trains a TransformerConv encoder using masked reconstruction, cross-modal alignment, and spatial smoothness objectives. On a private 11-sample melanoma cohort from an anonymized clinical collaborator comprising 54{,}912 total spots, LATTICE demonstrated stable optimization behavior, reproducible embeddings across analysis seeds, and complete multimodal integration across all samples. Adding scMultiome RNA to Visium RNA alone substantially improved concordance with Space Ranger clusters across 11 runs (adjusted Rand index [ARI] +0.157, normalized mutual information [NMI] +0.143, and spatial contiguity +0.174). Additional modalities further improved spatial contiguity and multimodal utility score (MUS), although they sometimes reduced agreement with RNA-derived reference labels, likely because the learned embeddings captured chromatin and regulatory structure beyond transcriptomic similarity alone. These results position LATTICE as a practical and empirically grounded framework for multimodal spatial omics integration, while also highlighting the need for stronger supervision and broader external benchmarking.
Submission history
From: Jagan Mohan Reddy Dwarampudi [
]
[v1]
Wed, 15 Jul 2026 22:54:47 UTC (3,779 KB)