Skip to content
DraconiticResearch
Search Sources API
201,818 studies31 sourcesUpdated 25 min ago
Peer-reviewedCC BY 4.0Nature Methods

Advancing morphological profiling with self-supervised learning through comparative insights into representation learning of cellular perturbations

  • Walid M. Abdelmoula1
  • Tomé Albuquerque2
  • Steffen Jaensch3
  • Pushpak Pati2
  • Michiel Van Dyck4
  • Samantha J. Allen5
  • Litao Zhang5
  • Tommaso Mansi6
  • Rui Liao6
6 affiliations
  1. 1Data Science and AI, Johnson & Johnson, Cambridge, MA USA
  2. 2Data Science and AI, Johnson & Johnson, Zug, Switzerland
  3. 3Discovery Technologies and Molecular Pharmacology, Johnson & Johnson, Beerse, Belgium
  4. 4Cognitive Automation and Technology, Johnson & Johnson, Beerse, Belgium
  5. 5Discovery Technologies and Molecular Pharmacology, Johnson & Johnson, Spring House, PA USA
  6. 6Data Science and AI, Johnson & Johnson, Titusville, NJ USA
Read full textPDFDOICite

Abstract

Self-supervised learning (SSL) enables scalable, label-free foundation models (FMs) that learn task-agnostic representations directly from images, bypassing cell segmentation and feature engineering inherent to classical morphological profiling. We present a large-scale evaluation of SSL FMs for image-based phenotypic profiling, covering contrastive (DINO), masked-based generative (MAE, iTPN) and hybrid (DINOv2) learning mechanisms, pretrained on all unannotated chemical perturbations from the Joint Undertaking for Morphological Profiling–Cell Painting (JUMP–CP) dataset (119,876 compounds, ~1.77 million images, 7 laboratories). Here we show that top-performing SSL models outperformed CellProfiler in compound-replicate retrieval on the held-out JUMP–Target2 set (302 annotated compounds, P ≤ 0.001). DINO and DINOv2 generalized across diverse perturbations and cell types, outperforming CellProfiler with 31–37% gains in site-averaged mean average precision for mechanism-of-action retrieval on EU-OPENSCREEN Bioactives (2,464 compounds), and consistently surpassing masked-based models across all benchmarks. This study provides a systematic assessment of SSL approaches for advancing morphological profiling and offers pretrained FMs to accelerate drug discovery.

Related

  1. scDRP: disentangled representation learning for predicting single-cell responses to perturbations and estimating individual treatment effects
    Briefings in bioinformaticsPeer-reviewedCC BYFull textMolecular biology & gene regulationEurope PMC
  2. Advancing Targeted Nanoparticle mRNA Delivery Using Machine Learning
    Marius Kausas, Valentina Schastlivaia
    ChemRxivPreprintCC BYFull textMolecular biology & gene regulationChemRxiv
  3. Encapsulated mitochondrial transplantation: Advancing organelle therapeutics
    Hae-Jun Yang, Pil-Soo Jeong, Sun-Uk Kim
    Signal Transduction and Targeted TherapyPeer-reviewedCC BY-NC-ND 4.0Full textMolecular biology & gene regulationPMCEurope PMC
  4. Editorial: CRISPR/CAS innovations: advancing precision and specificity in plant genetic engineering
    Giuseppe Dionisio, Siddharth Tiwari, Phanikanth Jogam
    Frontiers in Plant SciencePeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationCrossref
  5. Advancing asparaginase-based therapy for acute lymphoblastic leukemia through computational immunoengineering and molecular dynamics simulations
    Leila Sarafan Soleimanzadeh, Maryam Azimzadeh Irani, Bita Amiri Hazaveh +1
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textMolecular biology & gene regulationCrossref

Cite

BibTeXRIS
BibTeX
@article{abdelmoula2026advancing,
  title = {{Advancing morphological profiling with self-supervised learning through comparative insights into representation learning of cellular perturbations}},
  author = {Walid M. Abdelmoula and Tomé Albuquerque and Steffen Jaensch and Pushpak Pati and Michiel Van Dyck and Samantha J. Allen and Litao Zhang and Tommaso Mansi and Rui Liao},
  journal = {Nature Methods},
  year = {2026},
  month = oct,
  doi = {10.1038/s41592-026-03270-z},
  url = {https://doi.org/10.1038/s41592-026-03270-z},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Advancing morphological profiling with self-supervised learning through comparative insights into representation learning of cellular perturbations
AU  - Walid M. Abdelmoula
AU  - Tomé Albuquerque
AU  - Steffen Jaensch
AU  - Pushpak Pati
AU  - Michiel Van Dyck
AU  - Samantha J. Allen
AU  - Litao Zhang
AU  - Tommaso Mansi
AU  - Rui Liao
PY  - 2026
DA  - 2026/10/29
JO  - Nature Methods
DO  - 10.1038/s41592-026-03270-z
UR  - https://doi.org/10.1038/s41592-026-03270-z
AB  - Self-supervised learning (SSL) enables scalable, label-free foundation models (FMs) that learn task-agnostic representations directly from images, bypassing cell segmentation and feature engineering inherent to classical morphological profiling. We present a large-scale evaluation of SSL FMs for image-based phenotypic profiling, covering contrastive (DINO), masked-based generative (MAE, iTPN) and hybrid (DINOv2) learning mechanisms, pretrained on all unannotated chemical perturbations from the Joint Undertaking for Morphological Profiling–Cell Painting (JUMP–CP) dataset (119,876 compounds, ~1.77 million images, 7 laboratories). Here we show that top-performing SSL models outperformed CellProfiler in compound-replicate retrieval on the held-out JUMP–Target2 set (302 annotated compounds, P ≤ 0.001). DINO and DINOv2 generalized across diverse perturbations and cell types, outperforming CellProfiler with 31–37% gains in site-averaged mean average precision for mechanism-of-action retrieval on EU-OPENSCREEN Bioactives (2,464 compounds), and consistently surpassing masked-based models across all benchmarks. This study provides a systematic assessment of SSL approaches for advancing morphological profiling and offers pretrained FMs to accelerate drug discovery.
SN  - 1548-7091
SN  - 1548-7105
ER  -