aiqbench analysis
This research demonstrates how explicitly disentangling shared morphological features from modality-specific noise in spatial transcriptomics enhances downstream task performance compared to standard multimodal compression techniques. The study provides a critical evaluation framework for future foundation models aiming to integrate gene expression data with tissue morphology, highlighting that contrastive objectives generally outperform VAE-based approaches while offering nuanced gains through disentanglement strategies.
Disentangled Shared Representations Improve Morpho-Transcriptomic Integration
arXiv:2608.14355v1 Announce Type: new Abstract: Spatial transcriptomics (ST) enables the simultaneous profiling of gene expression and tissue morphology, creating an opportunity to learn multimodal representations capturing shared morpho-transcriptomic structure. However, standard multimodal models often compress modalities into a common latent space without explicitly separating shared and modality-specific sources of variation, which may limit downstream utility. We investigate whether explicit disentanglement of shared and private latent components improves multimodal representation learning for paired Hematoxylin \& Eosin (H\&E) and ST data. We compare VAE-based and contrastive approaches, each in standard and disentangled variants, across two cancer cohorts under matched experimental conditions. Representations are evaluated using cross-modal reconstruction, downstream probing and cross-modal probe transfer. The experiments suggest two main trends. First, contrastive objectives yield higher downstream probing performance than VAE-based models. Second, disentangled variants improve the selected reconstruction and probing metrics, although the gains depend on the model family, task, direction, and disentanglement strength. Overall, our results suggest that explicitly factorizing shared and modality-specific information can improve multimodal representation learning for spatial transcriptomics and provides a useful evaluation framework for future foundation models.
Abstract:Spatial transcriptomics (ST) enables the simultaneous profiling of gene expression and tissue morphology, creating an opportunity to learn multimodal representations capturing shared morpho-transcriptomic structure. However, standard multimodal models often compress modalities into a common latent space without explicitly separating shared and modality-specific sources of variation, which may limit downstream utility. We investigate whether explicit disentanglement of shared and private latent components improves multimodal representation learning for paired Hematoxylin \& Eosin (H\&E) and ST data. We compare VAE-based and contrastive approaches, each in standard and disentangled variants, across two cancer cohorts under matched experimental conditions. Representations are evaluated using cross-modal reconstruction, downstream probing and cross-modal probe transfer. The experiments suggest two main trends. First, contrastive objectives yield higher downstream probing performance than VAE-based models. Second, disentangled variants improve the selected reconstruction and probing metrics, although the gains depend on the model family, task, direction, and disentanglement strength. Overall, our results suggest that explicitly factorizing shared and modality-specific information can improve multimodal representation learning for spatial transcriptomics and provides a useful evaluation framework for future foundation models.
Artificial Intelligence (cs.AI)
arXiv:2608.14355 [cs.AI]
(or arXiv:2608.14355v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.14355 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 269 words
Related coverage
LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration
arXiv · Jul 21, 2026
CRAF: Cross-View Residual-Aware Fusion for Deepfake Speech Detection
arXiv · Sep 16, 2026
Learning Molecular Representations from Cellular Phenotypes with Structure Preservation
arXiv · Aug 5, 2026
Fine-Grained Instruction-Guided Graph Reasoning for Vision-and-Language Navigation
arXiv · Sep 9, 2026