Skip to content
DraconiticResearch
Search Sources API
226,742 studies31 sourcesUpdated 36 min ago
Peer-reviewedCC BYJournal of microscopy

Evaluating Integrative Strategies for Incorporating Phenotypic Features in Spatial Transcriptomics

  1. 1Imaging Platform, Broad Institute of MIT and Harvard, Cambridge MA, USA
  2. 2ETH Zurich, Zurich, Switzerland

Abstract

The key advantage of spatial transcriptomics (ST) technologies lies in the spatial domain: these techniques not only offer an unprecedented opportunity to interrogate intact biological samples in a spatially informed manner, but also set the stage for integration with other imaging-based modalities. However, how to most effectively exploit spatial context and integrate ST with imaging-based modalities that capture morphological insight remains an open and heavily investigated question. To address this, particularly under real-world experimental constraints such as limited dataset size, class imbalance, and bounding-box-based segmentation, we used a publicly available murine ileum Multiplexed Error-Robust Fluorescence In Situ Hybridisation (MERFISH) dataset to evaluate whether a minimally tuned variational autoencoder (VAE) could extract informative low-dimensional representations from cell crops of spot counts, nuclear stain, membrane stain, or a combination thereof. We assessed the resulting embeddings through PERMANOVA, cross-validated classification, and unsupervised clustering, and compared them to classical image-based feature vectors extracted via CellProfiler. While transcript counts (TC) generally outperformed other feature spaces, the VAE-derived latent spaces (LSs) captured meaningful biological variation and enabled improved label recovery for specific cell types. LS2, in particular, trained solely on morphological input, also exhibited moderate predictive power for a handful of genes in a ridge regression model. Notably, combining TC with LSs improved clustering outcomes in a strategy-dependent manner, yielding gains in either cluster homogeneity or label recovery under graph- or factor-based clustering approaches, respectively. In contrast, CellProfiler-derived features provided little benefit, highlighting the advantage of learned representations over handcrafted features. Collectively, these findings demonstrate that even under constrained conditions, VAEs can extract biologically meaningful signals from imaging data and constitute a promising strategy for multi-modal integration.

Related

  1. Integrative Transcriptomics and Machine Learning Nominate IL15- and PPP2R1A-Centered Programs in PTSD Using Pathway-Informed Autonomic–Cardiac Gene Prioritization
    Yihan Guo, Dongdong Shi, Lanying Liu +1
    International journal of molecular sciencesPeer-reviewedCC BYFull textGeneral biochemistryEurope PMC
  2. Dietary Interventions and Disease Outcomes in Finfish Feeding Trials Incorporating Host Transcriptomics: A Systematic Review and Meta-Analysis
    Luu Tang Phuc Khang, Sk Injamamul Islam, Onprang Sutthakiet +3
    International Journal of Molecular SciencesPeer-reviewedCC BY 4.0General biochemistryCrossref
  3. SpaMTP: integrative statistical analysis and visualization of spatial metabolomics and transcriptomics data
    Andrew Causer, Tianyao Lu, Jurgen Kriel +12
    Nature MethodsPeer-reviewedCC BY-NC-ND12 citedMolecular biology & gene regulationOpenAlex
  4. Enhancing cell type deconvolution in spatial transcriptomics through incorporating multiple reference panels
    Peiyao Zhao, Justin A Colacino, Ying Ma +1
    PLOS GeneticsPeer-reviewedCC BYFull textMolecular biology & gene regulationBioinformatics & computational biologyPLOSEurope PMCEurope PMCPMC
  5. Integrative Review on Tripterygium wilfordii Hook. f.: Chemical Constituents, Pharmacology, Toxicology, and Efficacy–Toxicity Modulation Strategies
    Jiaying Zhao, Chunyu Li, Lingli Mou +2
    International Journal of Molecular SciencesPeer-reviewedCC BY 4.0General biochemistryCrossref
BibTeX
@article{moser2026evaluating,
  title = {{Evaluating Integrative Strategies for Incorporating Phenotypic Features in Spatial Transcriptomics}},
  author = {Levin M Moser and Ahmad Kamal Hamid and Esteban Miglietta and Nodar Gogoberidze and Beth A Cimini},
  journal = {Journal of microscopy},
  year = {2026},
  month = jun,
  doi = {10.1111/jmi.70129},
  url = {https://doi.org/10.1111/jmi.70129},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Evaluating Integrative Strategies for Incorporating Phenotypic Features in Spatial Transcriptomics
AU  - Levin M Moser
AU  - Ahmad Kamal Hamid
AU  - Esteban Miglietta
AU  - Nodar Gogoberidze
AU  - Beth A Cimini
PY  - 2026
DA  - 2026/06/16
JO  - Journal of microscopy
DO  - 10.1111/jmi.70129
UR  - https://doi.org/10.1111/jmi.70129
AB  - The key advantage of spatial transcriptomics (ST) technologies lies in the spatial domain: these techniques not only offer an unprecedented opportunity to interrogate intact biological samples in a spatially informed manner, but also set the stage for integration with other imaging-based modalities. However, how to most effectively exploit spatial context and integrate ST with imaging-based modalities that capture morphological insight remains an open and heavily investigated question. To address this, particularly under real-world experimental constraints such as limited dataset size, class imbalance, and bounding-box-based segmentation, we used a publicly available murine ileum Multiplexed Error-Robust Fluorescence In Situ Hybridisation (MERFISH) dataset to evaluate whether a minimally tuned variational autoencoder (VAE) could extract informative low-dimensional representations from cell crops of spot counts, nuclear stain, membrane stain, or a combination thereof. We assessed the resulting embeddings through PERMANOVA, cross-validated classification, and unsupervised clustering, and compared them to classical image-based feature vectors extracted via CellProfiler. While transcript counts (TC) generally outperformed other feature spaces, the VAE-derived latent spaces (LSs) captured meaningful biological variation and enabled improved label recovery for specific cell types. LS2, in particular, trained solely on morphological input, also exhibited moderate predictive power for a handful of genes in a ridge regression model. Notably, combining TC with LSs improved clustering outcomes in a strategy-dependent manner, yielding gains in either cluster homogeneity or label recovery under graph- or factor-based clustering approaches, respectively. In contrast, CellProfiler-derived features provided little benefit, highlighting the advantage of learned representations over handcrafted features. Collectively, these findings demonstrate that even under constrained conditions, VAEs can extract biologically meaningful signals from imaging data and constitute a promising strategy for multi-modal integration.
SN  - 0022-2720
SN  - 1365-2818
ER  -