PRAGA: Prototype-aware Graph Adaptive Aggregation for Spatial Multi-modal Omics Analysis

TL;DR

A novel spatial multi-modal omics resolved framework, termed Prototype-aware Graph Adaptative Aggregation for Spatial Multi-modal Omics Analysis (PRAGA), which constructs a dynamic graph to capture latent semantic relations and comprehensively integrate spatial information and feature semantics.

Abstract

Spatial multi-modal omics technology, highlighted by Nature Methods as an advanced biological technique in 2023, plays a critical role in resolving biological regulatory processes with spatial context. Recently, graph neural networks based on K-nearest neighbor (KNN) graphs have gained prominence in spatial multi-modal omics methods due to their ability to model semantic relations between sequencing spots. However, the fixed KNN graph fails to capture the latent semantic relations hidden by the inevitable data perturbations during the biological sequencing process, resulting in the loss of semantic information. In addition, the common lack of spot annotation and class number priors in practice further hinders the optimization of spatial multi-modal omics models. Here, we propose a novel spatial multi-modal omics resolved framework, termed PRototype-Aware Graph Adaptative Aggregation for Spatial Multi-modal Omics Analysis (PRAGA). PRAGA constructs a dynamic graph to capture latent semantic relations and comprehensively integrate spatial information and feature semantics. The learnable graph structure can also denoise perturbations by learning cross-modal knowledge. Moreover, a dynamic prototype contrastive learning is proposed based on the dynamic adaptability of Bayesian Gaussian Mixture Models to optimize the multi-modal omics representations for unknown biological priors. Quantitative and qualitative experiments on simulated and real datasets with 7 competing methods demonstrate the superior performance of PRAGA. Code is available at https://github.com/Xubin-s-Lab/PRAGA.

Figures (4)

• Figure 1: Visualization of adjacency graph of random anchor spots on the Human Lymph Node dataset. (a) highlights the spots with the same cell type as the anchor spots. (b) and (c) shows the related spots found by spatial and feature adjacency K-nearest neighbor (KNN) graphs, respectively. (d) shows the potential relevant spots revealed by our proposed dynamic omic-specific graph.
• Figure 2: The framework of the proposed PRAGA. A learnable feature graph is used to explore the potential correlations between spots. The comprehensive representations of multi-modality (RNA, Protein) are obtained by aggregating modality-specific encodings, which GCNs calculate with linear combinations of learnable feature graphs and spatial adjacency graphs. The entire model is trained with a modality-specific reconstruction loss and a dynamic cluster prototype contrastive loss for the latent representation, where clusters are obtained via a Gaussian Mixture Model and optimized via split and merge operations.
• Figure 3: Visualization of qualitative experimental results on real datasets SPOTS mouse spleen (SPOTS) and mouse thymus stereo-CITE-seq (CITE).
• Figure 4: Effects of exponential moving average speed $\alpha$, temperature $\tau$ in $\mathcal{L}_{dpcl}$, and weight $\beta$ for $\mathcal{L}_{dpcl}$ on PRAGA performance measured by nine different metrics.