license: cc-by-nc-nd-4.0
language:
- en
tags:
- multiplex
- spatial-proteomics
- pathology
- vision
- pytorch
- self-supervised
- vit
extra_gated_prompt: >-
This model and associated code are released under the CC-BY-NC-ND 4.0 license
and may only be used for non-commercial, academic research purposes with
proper attribution. Any commercial use, sale, or other monetization of the
KRONOS model and its derivatives, which include models trained on outputs from
the KRONOS model or datasets created from the KRONOS model, is prohibited and
requires prior approval. Please note that the primary email used to sign up
for your Hugging Face account must match your institutional email to receive
approval. By downloading the model, you attest that all information
(affiliation, research use) is correct and up-to-date. Downloading the model
requires prior registration on Hugging Face and agreeing to the terms of use.
By downloading this model, you agree not to distribute, publish or reproduce a
copy of the model. If another user within your organization wishes to use the
KRONOS model, they must register as an individual user and agree to comply
with the terms of use. Users may not attempt to re-identify the deidentified
data used to develop the underlying model. If you are a commercial entity,
please contact the corresponding author.
extra_gated_fields:
Full name (first and last): text
Current affiliation (no abbreviations): text
Type of Affiliation:
type: select
options:
- Academia
- Industry
- label: Other
value: other
Current and official institutional email (**this must match your primary email in your Hugging Face account, @gmail/@hotmail/@qq email domains will be denied**): text
Please explain your intended research use: text
I agree to all terms outlined above: checkbox
I agree to use this model for non-commercial, academic purposes only: checkbox
I agree not to distribute the model, if another user within your organization wishes to use the TITAN model, they must register as an individual user: checkbox
metrics:
- accuracy
pipeline_tag: image-feature-extraction
KRONOS: A Foundation Model for Spatial Proteomics
KRONOSv2 model is COMING SOON! Please sign up to the waitlist by requesting access.
Abstract
Foundation models have transformed image analysis by acting as pretrained generalist backbones that can be adapted to many downstream tasks, with the recent advances in histopathology showing significant clinical promise. Their utility for spatial proteomics, the measurement of proteins at the subcellular resolution, has not been fully realized due to the lack of diverse, large-scale pretraining datasets and architectures to incorporate marker-specific subtleties. We introduce KRONOS, a foundation model built for spatial proteomics, trained in a self-supervised manner on over 56 million image patches covering 268 protein markers. KRONOS introduces key architectural innovations to account for heterogeneous marker panels that substantially vary across datasets, while modeling cross-marker interactions to better capture the underlying tissue biology. We demonstrate that KRONOS learns biologically meaningful representations across multiple scales from cells and tumor microenvironments to entire tissue, thereby enabling a wide range of spatial biology tasks. Evaluated comprehensively across diverse tasks such as cell phenotyping, region-level cell composition prediction, and patient risk stratification, KRONOS achieves consistently superior performance against raw marker-based approaches, which constitute the standard spatial proteomics analysis pipelines, and other image foundation models. Beyond benchmarking performance, KRONOS demonstrates new paradigms for data-driven clinical research workflow for biomarker discovery, supported by simultaneous superior risk stratification and interpretation of risk-modulating biomarkers, with the capabilities to "de-plex" high-dimensional discovery marker panels for effective downstream clinical deployment. We envision KRONOS as a foundational tool for data-driven, scalable spatial proteomics workflows.
Access
Access to this model is gated. To request access, click the Request access button on this page. Please submit your request from an institutional email address (e.g., a .edu address or an email from an affiliated academic or research institution). Requests are reviewed on a rolling basis, and you will be notified once access has been approved.
Contact
For any additional questions or comments, contact Andrew H. Song (asong2@mdanderson.org),
Anurag Vaidya (avaidya@mit.edu),
Sizun Jiang (sjiang3@bidmc.harvard.edu),
Faisal Mahmood (FaisalMahmood@bwh.harvard.edu).