Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

From modality-specific to compositional foundation models for cell biology.

Mojtaba Bahrami1, Till Richter2, Niklas A Schmacke3

  • 1Institute of Computational Biology, Helmholtz Center Munich, Munich, Germany; TUM School of Life Sciences Weihenstephan, Technical University of Munich, Munich, Germany.

Cell Systems
|February 19, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Continuous multi-omics pathway enrichment analysis resolves hidden functional heterogeneity.

Briefings in bioinformatics·2026
Same author

Transcriptional repression by TGIF2 coordinates neurogenic priming and neural stem cell maintenance.

Science advances·2026
Same author

CRISPR/Cas9-Mediated Gene Knockout Reveals a Nonredundant Role for p16<sup>INK4A</sup> in Controlling TCR-Dependent and Independent CD8 T Cell Expansion.

European journal of immunology·2026
Same author

UniversalEPI: robust prediction of cell type-specific and differential chromatin interactions from DNA sequence and chromatin accessibility.

Nucleic acids research·2026
Same author

RegVelo: Gene-regulatory-informed dynamics of single cells.

Cell·2026
Same author

Glial multicellular programs reveal distinct patient stratification in Parkinson's disease.

Research square·2026
Same journal

High-throughput DNA engineering by mating bacteria.

Cell systems·2026
Same journal

Living bacterial reservoir computers for information processing and sensing.

Cell systems·2026
Same journal

A data-driven modeling framework for mapping genotypes to synthetic microbial community functions.

Cell systems·2026
Same journal

BulkFormer: A large-scale foundation model for bulk transcriptomes.

Cell systems·2026
Same journal

Glycoform engineering of a mammalian platform to sculpt a humanized recombinant bioscavenger.

Cell systems·2026
Same journal

Targeted genomic editing of human gut Bacteroides species based on CRISPR-associated transposases.

Cell systems·2026
See all related articles
This summary is machine-generated.

Compositional AI offers a modular approach to multimodal foundation models, unifying diverse cell biology data for a deeper understanding of cellular states in health and disease.

Area of Science:

  • Computational Biology
  • Artificial Intelligence in Biology
  • Systems Biology

Background:

  • Understanding cellular states in health and disease requires integrating diverse single-cell measurements.
  • Current multimodal models face challenges in learning generalizable representations across different data types and biological contexts.

Purpose of the Study:

  • To explore compositional AI as a modular design for multimodal foundation models in cell biology.
  • To unify various biological data modalities into cohesive representations of cellular behavior.
  • To address challenges in data availability and modality representation for advanced AI models.

Main Methods:

  • Utilizing compositional AI and deep learning for multimodal foundation models.
  • Employing transformer-based attention strategies to integrate diverse biological data.
Keywords:
compositional AIsingle-cell foundation modelssingle-cell multi-omics

Related Experiment Videos

  • Developing methods for connecting and aligning partially overlapping multimodal measurements.

    • Main Results:

    • Compositional AI provides a framework for unifying modalities like genomics, proteomics, and imaging.
    • Transformer attention mechanisms can effectively handle structural differences and limited data.
    • A comprehensive representation space can be built by aligning diverse biological data.

    Conclusions:

    • Compositional AI is a promising approach for building robust multimodal foundation models in cell biology.
    • These models can lead to agentic virtual cell models for decoding cellular complexity.
    • Future directions include further development for broader applications in biological research.