Resources: Posters

Multi-modal Single-Cell Analysis: Self-supervised Vision Models Reveal Hidden Morphology-Transcriptome Relationships Across Diverse Biological Systems

November 12, 2025

Cold Spring Harbor Labs - Single Cell Analyses

Shreya Deshmukh, Ph.D., Sr. Data Scientist

Keywords: R3200 Platform, CellCage™ technology, Multi-modal analysis, Self-supervised learning, Computer vision, Morphology-transcriptomics, Single-cell AI

‍

View Poster

This research demonstrates the power of self-supervised vision models in uncovering complex, non-linear relationships between cell morphology and transcriptomic profiles. By utilizing the R3200 Platform and CellCage™ technology to maintain spatial and temporal context across diverse biological systems, the study identifies hidden cellular states that traditional analysis methods overlook. The results establish a robust framework for linking visual phenotypes to gene expression at single-cell resolution.

Back to Posters

Case Study: Functional Profiling of Microglia in Neuroinflammation

Link microglial behavior to gene expression at single-cell resolution, for insight into neuroinflammation, drug response, and immune dysfunction in CNS disease.

View Case Study Details

Phagocytosis assay in CellCage enclosures - (click to expand image).

Researchers used the Cellanome R3200 to enclose individual microglia with fluorescent particles and track phagocytosis over 12 hours via fluorescent imaging. Each cell’s transcriptome was then sequenced, linking activity levels to gene expression.

What they found:

High-activity microglia upregulated genes in complement signaling, lipid metabolism, and lysosomal function–key pathways in neuroinflammation and repair.

Why it matters:

This approach overcomes key limitations in standard assays by capturing phagocytic function and gene expression in the same individual cells without dissociation, pooling, or inference. It enables a direct, scalable readout of immune heterogeneity, and reveals the transcriptional programs driving effective or impaired microglial responses.

‍What’s next:

Extend to co-cultures by layering enclosed microglia over intact neuronal networks. Study how cell-cell interactions shape phagocytic behavior and fate. Combine with cytokines, CRISPR libraries, or immunotherapies to generate time-resolved, multi-modal datasets that can be used for MoA analysis, early biomarker discovery, and AI-guided modeling in CNS disease.

View Poster

Case Study: Modeling Synapse Formation and Developmental Trajectories in 3D

Track development, function, and gene expression in intact neurospheres, a human-relevant 3D model increasingly vital as regulators move away from animal studies.

View Case Study Details

Neurospheres in windowed cages that allow them to reach out to form a synapse - (click to expand image).

Stem-cell-derived neurospheres offer a robust 3D model of early brain development, but standard assays disrupt their structure and miss critical dynamics.

Approach:

Using the Cellanome R3200, the research team explored,

Hundreds of intact neurospheres (100–200 cells each) were cultured inside individual CellCage™ enclosures.
Axon extension, synapse formation and calcium activity were tracked over multiple days.
End-point RNA-Seq was linked back to each neurosphere’s functional behavior.
UMAP clustering revealed lineage-specific gene programs, validated by fluorescent markers.

‍What's next:

This lays the groundwork for CRISPR-based multimodal screens to probe mechanisms of development, degeneration, and repair within preserved 3D architecture.

‍Why it matters:

As the FDA and others move to reduce reliance on animal models, human-relevant in vitro systems like this are increasingly essential.

View Poster

View Poster Presentation

FAQ's

How are self-supervised vision models applied on the R3200 Platform?

Self-supervised vision models are AI algorithms trained to recognize patterns in cell morphology without manual labeling. On the R3200 Platform, these models process high-resolution live-cell imaging data to extract high-dimensional morphological features that are then directly correlated with the cell's transcriptomic profile.

How does CellCage™ technology help reveal hidden morphology-transcriptome relationships?

CellCage™ technology secures individual cells in place, allowing the R3200 Platform to capture a continuous visual record of a cell's physical appearance before sequencing. This preservation of context allows researchers to identify subtle structural changes that correspond to specific gene expression programs across various cell types.

Why is multi-modal analysis across diverse biological systems important for this study?

Cellular functions are governed by both physical form and genetic activity. Analyzing these relationships across different systems ensures that the morphology-transcriptome links identified by the R3200 Platform are biologically robust, providing a more holistic understanding of cell identity and state transitions.

Explore Applications

See a Demo