CM4AI Dataset Documentation

Description	ID	Name	Response
Create AI-ready maps of human cell architecture from disease-relevant cell lines as part of the NIH ...	cm4ai:purpose:ai-ready-maps	Create AI-ready cell architecture maps	CM4AI's overarching mission is to produce ethical, AI-ready datasets of cell architecture, inferred ...
Enable interpretable genotype-phenotype learning by providing visible machine learning systems infor...	cm4ai:purpose:interpretable-genotype-phenotype	Enable interpretable genotype-phenotype learning	The CM4AI project seeks to map the spatiotemporal architecture of human cells and use these maps tow...

Description	ID	Instance Type	Name
Expressed genome-scale CRISPRi Perturbation Cell Atlas in undifferentiated KOLF2.1J human induced pl...	cm4ai:instance:crispr-perturbation	Perturb-seq single-cell RNA sequencing data	CRISPR Perturbation Cell Atlas
Size exclusion chromatography-mass spectroscopy (SEC-MS) data from undifferentiated KOLF2.1J human i...	cm4ai:instance:sec-ms-ipsc	Protein-protein interaction mass spectrometry data	SEC-MS protein-protein interactions in iPSCs
SEC-MS protein-protein interaction data from iPSC-derived NPCs (neural progenitor cells), neurons, a...	cm4ai:instance:sec-ms-differentiated	Protein-protein interaction mass spectrometry data	SEC-MS in differentiated iPSC derivatives
Spatial localization of 563 proteins of interest in untreated cells of the breast cancer cell line M...	cm4ai:instance:if-images-untreated	Immunofluorescence confocal microscopy images	Immunofluorescence images - untreated MDA-MB-468
Spatial localization of 563 proteins in MDA-MB-468 breast cancer cells treated with paclitaxel chemo...	cm4ai:instance:if-images-paclitaxel	Immunofluorescence confocal microscopy images	Immunofluorescence images - paclitaxel-treated MDA-MB-468
Spatial localization of 563 proteins in MDA-MB-468 breast cancer cells treated with vorinostat chemo...	cm4ai:instance:if-images-vorinostat	Immunofluorescence confocal microscopy images	Immunofluorescence images - vorinostat-treated MDA-MB-468
AP-MS on endogenously tagged cell lines for protein-protein interaction mapping. 17 genes endogenous...	cm4ai:instance:ap-ms	Protein-protein interaction mass spectrometry data	Affinity purification mass spectrometry
Hierarchical directed acyclic graph (DAG) maps of cell architecture produced by integrating protein ...	cm4ai:instance:cell-maps	Network data structure (hierarchical graph)	Hierarchical cell maps

Access Urls	Description	ID	Name
https://dataverse.lib.virginia.edu/dataset.xhtml?persistentId=doi:10.18130/V3/B35XWX	Datasets packaged in RO-Crate format with rich metadata and provenance graphs.	cm4ai:format:ro-crate	RO-Crate packages
https://dataverse.lib.virginia.edu	University of Virginia's LibraData (Dataverse) NIH-approved generalist repository.	cm4ai:format:dataverse	LibraData Dataverse repository
https://cm4ai.org, http://www.cm4ai.org	CM4AI project portal using U-BRITE platform for open dissemination.	cm4ai:format:portal	CM4AI web portal
https://www.ndexbio.org	Cell maps shared via NDEx for visualization in web browsers or Cytoscape.	cm4ai:format:ndex	Network Data Exchange

🔍

Collection Process

How was the data acquired?

ID *

https://doi.org/10.18130/V3/B35XWX

Name

CM4AI

Title

Cell Maps for Artificial Intelligence - March 2025 Data Release (Beta)

Description

This dataset is the March 2025 Data Release of Cell Maps for Artificial Intelligence (CM4AI; CM4AI.org), the Functional Genomics Grand Challenge in the NIH Bridge2AI program. This Beta release includes perturb-seq data in undifferentiated KOLF2.1J iPSCs; SEC-MS data in undifferentiated KOLF2.1J iPSCs and iPSC-derived NPCs, neurons, and cardiomyocytes; and IF images in MDA-MB-468 breast cancer cells in the presence and absence of chemotherapy (vorinostat and paclitaxel). CM4AI's objective is to deliver machine-readable hierarchical maps of cell architecture as AI-Ready data produced from multimodal interrogation of 100 chromatin modifiers and 100 metabolic enzymes involved in cancer, neuropsychiatric, and cardiac disorders in disease-relevant cell lines under perturbed and unperturbed conditions. CM4AI output data are packaged with provenance graphs and rich metadata as AI-ready datasets in RO-Crate format using the FAIRSCAPE framework. CM4AI is a collaboration of UCSD, UCSF, Stanford, UVA, Yale, UA Birmingham, Simon Fraser University, and the Hastings Center.

Page

https://dataverse.lib.virginia.edu/dataset.xhtml?persistentId=doi:10.18130/V3/B35XWX

Language

Keywords

AI
affinity purification
AP-MS
artificial intelligence
breast cancer
Bridge2AI
cardiomyocyte
CM4AI
CRISPR/Cas9
CRISPR/Cas perturbation screens
functional genomics
induced pluripotent stem cell
iPSC
KOLF2.1J
machine learning
mass spectroscopy
MDA-MB-468
neural progenitor cell
NPC
neuron
paclitaxel
perturb-seq
perturbation sequencing
protein-protein interaction
protein localization
single-cell RNA sequencing
scRNAseq
SEC-MS
size exclusion chromatography
subcellular imaging
vorinostat
RO-Crate
FAIRSCAPE
cell maps
hierarchical maps
visible machine learning
VNN
visible neural networks
chromatin modifiers
metabolic enzymes
immunofluorescence
IF imaging
spatial proteomics

Addressing Gaps

Description	ID	Name	Response
Address the limitation of black-box machine learning models in genomics by creating visible neural n...	cm4ai:gap:black-box-ml	Address black-box machine learning	Machine learning models show great promise in analyzing the human genome but their inner workings ar...
Fill the gap in availability of high-quality, FAIR, AI-ready functional genomics datasets with rich ...	cm4ai:gap:fair-ai-datasets	Provide FAIR AI-ready functional genomics datasets	CM4AI addresses the need for ethical, FAIR, AI-Ready biomedical data that are fully characterized wi...

Creators

Role	Name	ORCID	Affiliation
Contributor	Trey Ideker	cm4ai:creator:ideker	-
Contributor	Timothy W Clark	cm4ai:creator:clark	-
Contributor	Nevan J Krogan	cm4ai:creator:krogan	-
Contributor	Emma Lundberg	cm4ai:creator:lundberg	-
Contributor	Prashant Mali	cm4ai:creator:mali	-
Contributor	Andrej Sali	cm4ai:creator:sali	-
Contributor	Jean-Christophe Bélisle-Pipon	cm4ai:creator:belisle-pipon	-
Contributor	Cynthia Brandt	cm4ai:creator:brandt	-
Contributor	Jake Yue Chen	cm4ai:creator:chen	-
Contributor	Ying Ding	cm4ai:creator:ding	-
Contributor	Samah Fodeh	cm4ai:creator:fodeh	-
Contributor	Pamela Payne-Foster	cm4ai:creator:payne-foster	-
Contributor	Sarah J Ratcliffe	cm4ai:creator:ratcliffe	-
Contributor	Vardit Ravitsky	cm4ai:creator:ravitsky	-
Contributor	Wade Loren Schulz	cm4ai:creator:schulz	-

Subsets

Description	ID	Name
Expressed genome-scale CRISPRi Perturbation Cell Atlas in undifferentiated KOLF2.1J human induced pl...	cm4ai:subset:crispr-atlas	CRISPR Perturbation Cell Atlas
Size exclusion chromatography-mass spectroscopy data on undifferentiated KOLF2.1J human induced plur...	cm4ai:subset:sec-ms-ipsc	Protein-protein Interaction SEC-MS iPSCs
Spatial localization of 563 proteins in untreated MDA-MB-468 breast cancer cells imaged by ICC-IF an...	cm4ai:subset:if-untreated	Protein Localization Images - Untreated
Spatial localization of 563 proteins in MDA-MB-468 breast cancer cells treated with paclitaxel, imag...	cm4ai:subset:if-paclitaxel	Protein Localization Images - Paclitaxel
Spatial localization of 563 proteins in MDA-MB-468 breast cancer cells treated with vorinostat, imag...	cm4ai:subset:if-vorinostat	Protein Localization Images - Vorinostat

Sampling Strategies

Description	ID	Name
Disease-relevant cell lines selected to enable AI-ready functional genomics mapping for cancer, neur...	cm4ai:sampling:cell-line-selection	Disease-relevant cell line selection
Selection of 100 chromatin regulators and 100 metabolic enzymes involved in cancer, neuropsychiatric...	cm4ai:sampling:chromatin-modifiers	Chromatin modifier selection
Selection of chemotherapy treatments (paclitaxel and vorinostat) and cell differentiation states to ...	cm4ai:sampling:perturbation-conditions	Perturbation and treatment conditions

External Resources

Description	External Resources	ID	Name
Official CM4AI project website and portal	https://cm4ai.org, https://www.cm4ai.org	cm4ai:resource:project-website	CM4AI Project Website
Key publications describing CM4AI methods and datasets	https://doi.org/10.1101/2024.05.21.589311, https://doi.org/10.1101/2024.11.03.621734	cm4ai:resource:publications	CM4AI Publications
Platform for sharing and visualizing cell maps	https://www.ndexbio.org	cm4ai:resource:ndex	Network Data Exchange (NDEx)
AI-readiness framework for metadata and provenance	https://fairscape.github.io	cm4ai:resource:fairscape	FAIRSCAPE Framework
Software tools for cell map creation and analysis	https://github.com/idekerlab, http://integrativemodeling.org	cm4ai:resource:software-tools	CM4AI Software Tools

Acquisition Methods

Acquisition Details	Description	ID	Name	Was Directly Observed
Endogenous tagging of genes in MDA-MB-468 cells, AP-MS data acquisition under three conditions (untreated, paclitaxel, vorinostat), 17 genes tagged in Year 1, 34 additional genes in process	AP-MS on endogenously tagged cell lines to map protein-protein interactions of 100 chromatin regulat...	cm4ai:acquisition:ap-ms	Affinity purification mass spectrometry	True
SEC-MS performed on MDA-MB-468 cells under three conditions, Identified 72/100 chromatin modifiers with 52 in protein complexes, ... (+3 more)	SEC-MS for proteome-wide complex and interaction mapping as orthogonal approach to AP-MS.	cm4ai:acquisition:sec-ms	Size exclusion chromatography mass spectrometry	True
Automated fixation and permeabilization protocols using pipetting robot, Four-channel staining (DAPI, calreticulin, tubulin, target protein), ... (+3 more)	IF staining and high-resolution confocal microscopy for spatial proteomics mapping of protein subcel...	cm4ai:acquisition:immunofluorescence	Immunofluorescence confocal microscopy	True
CRISPR lentiviral library targeting 100 chromatin factors with 6 guide RNAs per gene, MDA-MB-468 and KOLF2 CRISPR lines with inducible dCas9, ... (+4 more)	CRISPR-Cas9 perturbation screens for genetic perturbation mapping with single-cell RNA sequencing re...	cm4ai:acquisition:crispr-perturbation	Single-cell CRISPR perturbation screens	True

Collection Mechanisms

Description	ID	Mechanism Details	Name
Mass spectrometry instruments for AP-MS and SEC-MS protein interaction mapping.	cm4ai:mechanism:mass-spectrometry	Affinity purification coupled with mass spectrometry, Size exclusion chromatography coupled with mass spectrometry, Proteome-wide interaction profiling	Mass spectrometry instruments
High-resolution confocal microscopy for immunofluorescence imaging of protein subcellular localizati...	cm4ai:mechanism:confocal-microscopy	Confocal microscopy with four-channel imaging, Immunofluorescence-based staining (ICC-IF), Automated pipetting robot for sample preparation	Confocal microscopy
CRISPR-Cas9 gene perturbation with single-cell RNA sequencing for transcriptional and fitness phenot...	cm4ai:mechanism:crispr-cas9	CRISPRi gene knockdown with inducible dCas9, Lentiviral library delivery, ... (+2 more)	CRISPR-Cas9 perturbation system
Multi-Scale Integrated Cell (MuSIC) pipeline for integrating multimodal data streams into hierarchic...	cm4ai:mechanism:music-pipeline	node2vec deep learning for PPI network embedding, Human Protein Atlas deep learning for image embedding, ... (+4 more)	MuSIC data integration pipeline

Data Collectors

Collector Details	Description	ID	Name
Ideker Lab - project coordination and cell map integration, Mali Lab - genetic perturbation mapping, Sali Lab - integrative structure modeling	UCSD teams led by Trey Ideker (PI), Prashant Mali, and Andrej Sali.	cm4ai:collector:ucsd	University of California San Diego
Krogan Laboratory - AP-MS and SEC-MS data generation, Protein complex mapping across conditions	UCSF teams led by Nevan Krogan for protein-protein interaction data.	cm4ai:collector:ucsf	University of California San Francisco
Lundberg Laboratory - immunofluorescence imaging, Spatial proteomics mapping with confocal microscopy, Copyright holder for raw image data	Stanford team led by Emma Lundberg for spatial proteomics.	cm4ai:collector:stanford	Stanford University
Standards Module - FAIRSCAPE framework development, AI-readiness packaging and metadata, Data repository (LibraData/Dataverse)	UVA team led by Timothy Clark for AI-readiness standards.	cm4ai:collector:uva	University of Virginia
Data curation and standards, Internship program hosting	Yale team led by Cynthia Brandt, Samah Fodeh, and Wade Schulz.	cm4ai:collector:yale	Yale University
Teaming Module - U-BRITE platform for portal, Open dissemination of data, maps, and tools, Internship program hosting	UAB team led by Jake Chen for teaming and dissemination.	cm4ai:collector:uab	University of Alabama at Birmingham

Collection Timeframes

Description	ID	Name	Timeframe Details
September 1, 2022 to August 31, 2026 with ongoing data augmentation.	cm4ai:timeframe:project-period	Overall project period	Project start September 1 2022, Project end August 31 2026, ... (+4 more)
Year 1 data generation including chromatin modifiers, spatial proteomics, and CRISPR screens.	cm4ai:timeframe:year1	Year 1 accomplishments	100 chromatin regulators mapped in MDA-MB-468 cells, Single-cell CRISPR screens completed, ... (+2 more)

Preprocessing Strategies

Description	ID	Name	Preprocessing Details
Generation of embeddings from PPI networks and IF images using deep learning models for dimensionali...	cm4ai:preprocessing:embedding	Deep learning embeddings	node2vec deep learning for PPI network embedding, Human Protein Atlas deep learning for image embedding, Contrastive learning for co-embedding integration
Integration of PPI and image embeddings through MuSIC pipeline to create co-embeddings and hierarchi...	cm4ai:preprocessing:integration	Multimodal data integration	Co-embedding integration with minimal information loss, Community detection based on co-embedding similarities, Hierarchy creation from multi-resolution community detection
Annotation of protein assemblies in cell maps using ontology alignment and large language models.	cm4ai:preprocessing:annotation	Cell map annotation	Alignment to Gene Ontology and Reactome pathways, LLM-based naming of protein assemblies with confidence scores, Evaluation of overlap with known cell biology
Packaging of datasets with rich metadata, provenance graphs, and schemas in RO-Crate format for AI-r...	cm4ai:preprocessing:ro-crate-packaging	RO-Crate packaging with FAIRSCAPE	FAIRSCAPE-CLI client for RO-Crate creation, JSON-Schema data dictionaries for all datasets, ... (+4 more)

Maintainers

ID
cm4ai:maintainer:cm4ai-team
Name
CM4AI Project Team
Description
Multi-institutional team responsible for dataset maintenance.
Maintainer Details
- Point of contact Trey Ideker (University of California San Diego)
- Data depositor Justin Niestroy (University of Virginia)
- Data Access Committee for ethical supervision

Ethical Reviews

Description	ID	Name	Review Details
Ethics team employs Value-Sensitive Design (VSD) methodology for ethical AI development.	cm4ai:ethics:value-sensitive-design	Value-Sensitive Design framework	Conceptual work to craft axiological repository of values, Design standards and expectations for responsible AI, ... (+2 more)
Mixed empirical research combining qualitative and quantitative methods to capture community insight...	cm4ai:ethics:stakeholder-engagement	Stakeholder engagement research	Diverse perspective integration in ethical guidelines, Community insights on values and concerns, Tools for ethical, legal, and social ramifications awareness
Bridge2AI Data Sharing and Dissemination Working Group Code of Conduct for data access.	cm4ai:ethics:code-of-conduct	Bridge2AI Code of Conduct	Basic requirements for Bridge2AI Open House participants, Attestation required prior to accessing CM4AI datasets, Ethical commitments from data users

Data Protection Impacts

Description	ID	Impact Details	Name
Ethical preparation, licensing, dissemination, and data access supervision balancing openness with i...	cm4ai:protection:licensing	CC BY-NC-SA 4.0 for open non-commercial use, Commercial use requires separate licensing, ... (+2 more)	Licensing and data governance
Guidelines and best practices for ethical development of AI systems leveraging CM4AI data.	cm4ai:protection:responsible-ai	Anticipation of potential AI/ML applications, Integration into data governance frameworks, ... (+2 more)	Responsible AI development framework

🚀

Uses

What (other) tasks could the dataset be used for?

Tasks

Description	ID	Name	Response
Create hierarchical directed acyclic graph (DAG) maps where each node represents an assembly of prot...	cm4ai:task:cell-mapping	Hierarchical cell mapping	Cell maps are hierarchical directed acyclic graphs (DAG), where each node represents an assembly of ...
Integrate three complementary mapping approaches - proteomic mass spectrometry, cellular imaging, an...	cm4ai:task:multimodal-integration	Multimodal data integration	The project launches a coordinated effort involving three complementary mapping approaches: affinity...
Package datasets with FAIR compliance, provenance graphs, complete schemas, validation procedures, d...	cm4ai:task:ai-readiness	AI-readiness packaging	AI-Ready biomedical data are fully characterized FAIR data of known provenance, which can be ethical...

Existing Uses

Description	Examples	ID	Name
Scientific publications describing CM4AI methods and initial findings.	Clark T, et al. Cell Maps for Artificial Intelligence. bioRxiv 2024.05.21.589311, Nourreddine S, et al. A PERTURBATION CELL ATLAS OF HUMAN iPSCs. bioRxiv 2024.11.03.621734	cm4ai:use:publications	Publications using CM4AI data
March 2024 CodeFest event for training researchers in CM4AI data and tools.	38 registrants with 40% female and 30% from underrepresented communities, Detailed tutorials on data derivation and integration, Production-ready software demonstration	cm4ai:use:codefest	CM4AI CodeFest training event

Intended Uses

Description	Examples	ID	Name
Development of visible neural networks (VNNs) that provide interpretable AI/ML predictions based on ...	Visible machine learning systems informed by cell maps, Interpretable genotype-phenotype predictions, ... (+2 more)	cm4ai:intended:visible-ml	Visible machine learning development
AI/ML research on functional genomics data including transcriptional phenotypes, protein interaction...	Machine learning on multimodal functional genomics data, Disease mechanism understanding, ... (+2 more)	cm4ai:intended:biomedical-ai	Biomedical AI research
Research on protein assemblies, cellular compartments, and protein complex organization.	Protein complex characterization, Subcellular localization studies, ... (+2 more)	cm4ai:intended:cell-biology	Cell biology research

Future Use Impacts

Description	ID	Impact Details	Name
Dataset characteristics specific to MDA-MB-468 and KOLF2.1J cell lines may not generalize to all cel...	cm4ai:impact:cell-line-specific	MDA-MB-468 specific to triple-negative breast cancer context, KOLF2.1J from healthy male Northern European donor, ... (+2 more)	Cell line-specific characteristics
Limitations inherent to mass spectrometry, microscopy, and CRISPR perturbation technologies.	cm4ai:impact:technological-limitations	Mass spectrometry detection limits, Antibody availability and specificity constraints, ... (+3 more)	Technological and methodological limitations

Discouraged Uses

Description	Discouragement Details	ID	Name
Commercial use is restricted by CC BY-NC-SA 4.0 license and requires separate licensing negotiation....	Commercial use not permitted under CC BY-NC-SA 4.0, Separate license required from copyright holders (UCSD, Stanford, UCSF), Data Access Committee supervision for commercial licensing	cm4ai:discouraged:commercial	Commercial use without license
Data are from cell lines, not clinical samples, and should not be directly applied to clinical decis...	Non-clinical data from tissue cultures, De-identified and cannot be matched to human subjects, Requires clinical validation before medical application	cm4ai:discouraged:clinical-direct	Direct clinical application without validation

License And Use Terms

ID

cm4ai:license:cc-by-nc-sa

Name

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International

Description

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International license. Data are Copyright 2025 The Regents of the University of California except where otherwise noted. Spatial proteomics raw image data copyright 2025 The Board of Trustees of the Leland Stanford Junior University.

License Terms

Attribution required to copyright holders and authors
Citation of bioRxiv article required
Non-commercial use only without separate license
ShareAlike - derivatives must use same license
Software under BSD-3 and MIT open source licenses
FAIRSCAPE under MIT license

Motivation

Composition

Collection Process

Uses

Distribution

Maintenance

Human Subjects