Rubric20-Semantic Evaluation Report

All mandatory fields present with exceptional detail. Description is comprehensive (450+ characters), keywords are extensive (26 terms), license terms include specific citation requirements and Data Access Committee oversight.

id: https://doi.org/10.18130/V3/DXWOS5, title: Cell Maps for Artificial Intelligence (CM4AI), description: 450+ chars, keywords: 26 keywords, license_and_use_terms: comprehensive CC BY-NC-SA 4.0 with detailed attribution requirements

DOI format correct (10.18130 = Harvard Dataverse). All fields semantically appropriate for functional genomics dataset.

Excellent narrative depth. Description provides comprehensive technical detail including data modalities (IF, AP-MS, SEC-MS, CRISPR), target proteins (100 chromatin modifiers, 100 metabolic enzymes), cell lines, and MuSIC pipeline methodology.

description: 450+ chars, purposes: 3 detailed entries averaging 250+ chars each, tasks: 4 entries averaging 200+ chars

Narrative content is technically accurate and domain-appropriate. Terminology (AP-MS, SEC-MS, CRISPR, MuSIC, FAIRSCAPE) used correctly.

Exceptional keyword coverage spanning project identity, methodologies, biological systems, cell lines, data standards, and technical approaches. Keywords are domain-specific and appropriate for genomics/proteomics research.

keywords: 26 unique terms including Cell Maps, AI-Ready Data, Bridge2AI, Protein-Protein Interactions, Spatial Proteomics, CRISPR Perturbation, MDA-MB-468, KOLF2.1J, AP-MS, SEC-MS, FAIR Principles, RO-Crate, FAIRSCAPE

Keywords semantically appropriate and well-balanced across: project identity (CM4AI, Bridge2AI), methods (AP-MS, SEC-MS, CRISPR), biological context (chromatin, metabolism, breast cancer), and standards (FAIR, RO-Crate).

Excellent format diversity reflecting multi-modal data: structured metadata (RO-Crate/JSON-LD), raw experimental data (mass spec, sequencing), processed results (cell maps), and archival packages (Dataverse). Each format serves distinct purpose in data ecosystem.

distribution_formats: 5 distinct formats documented - RO-Crate packages (JSON-LD metadata), Mass Spectrometry data (MassIVE), Sequence data (NCBI SRA), Hierarchical Cell Maps (NDEx networks), Dataverse archives

Format choices are semantically appropriate: RO-Crate for AI-ready packaging, MassIVE for proteomics community standards, SRA for genomics standards, NDEx for network visualization.

Comprehensive quantification across all data modalities with specific counts for proteins analyzed, genes targeted, complexes detected, and experimental conditions.

instances: 4 detailed instance definitions (MDA-MB-468, KOLF2.1J, 100 chromatin regulators, 100 metabolic enzymes). Subsets documented: 563 proteins in IF images, 17 genes tagged for AP-MS (34 in progress), 72/100 chromatin modifiers in SEC-MS, 11,739 targeted genes in CRISPR screens, 1,000+ protein complexes detected in MDA-MB-468, 700+ in iPSCs

Instance counts are realistic and internally consistent. Numbers align with project scope (100+100 target proteins, multiple experimental replicates across conditions).

Excellent identifier coverage with multiple persistent IDs: primary dataset DOI, version-specific DOIs, publication DOI, project URL, and cell line RRIDs from Cellosaurus.

id: https://doi.org/10.18130/V3/DXWOS5, page: https://www.cm4ai.org, additional DOIs: 10.18130/V3/B35XWX (v1.4), 10.18130/V3/F3TD5R (v2.1), bioRxiv DOI: 10.1101/2024.05.21.589311, RRIDs: CVCL_0419 (MDA-MB-468), CVCL_B5P3 (KOLF2.1J)

All identifier formats validated: DOI prefix 10.18130 is Harvard Dataverse (correct), RRID format matches Cellosaurus cell line registry (correct), bioRxiv DOI format correct.

Exceptional funding documentation with complete grant numbers, budget details, project timeline, and comprehensive creator list with institutional affiliations and module leadership roles.

funders: NIH Common Fund Bridge2AI grant 1OT2OD032742-01 (Bridge2AI Functional Genomics) and 5U54HG012513-02 (Bridge2AI Bridge Center), Frederick Thomas Fund. creators: 15 investigators with names, roles, and institutional affiliations (UCSD, UCSF, Stanford, UVA, Yale, UAB, SFU, UMontreal, UT Austin). Funding details include opportunity number OTA-21-008, project dates Sept 2022-Aug 2026, FY2025 budget $5,289,382

Grant number 1OT2OD032742-01 follows NIH format correctly (OT2 = Other Transaction for Research, OD = Office of Director). Budget amount realistic for NIH Bridge2AI flagship project. Institutional affiliations match known Bridge2AI participants.

Strong ethics documentation appropriate for non-human subjects research. Clearly explains why IRB not required (commercially available de-identified cell lines), documents ethics oversight (Ethics Module, Data Access Committee), acknowledges donor demographics while explaining non-identifiability. Score 4/5 because full human subjects protections (consent, compensation, privacy) not applicable to this research type.

human_subject_research.involves_human_subjects: false with detailed justification (de-identified commercial cell lines, cannot be matched to individuals), irb_approval: 'Not applicable - de-identified cell lines', ethics_review_board: CM4AI Ethics Module (Ravitsky, Bélisle-Pipon), Data Access Committee (Parker), special_populations: donor demographics documented (MDA-MB-468: 51yo Black female, KOLF2.1J: male Northern European), sensitive_elements.sensitive_elements_present: false with explanation

Non-human subjects determination is semantically correct for ATCC and HipSci cell lines. Ethics governance appropriate with Data Access Committee for distribution oversight. Cell line sourcing (ATCC, HipSci) validates ethical procurement claims.

Excellent governance documentation with clear license terms, IP ownership, commercial restrictions, attribution requirements, and oversight mechanisms. Addresses both open science goals and IP protection needs.

license: CC BY-NC-SA 4.0, license_and_use_terms: comprehensive description with attribution requirements, commercial use restrictions ('Commercial use requires separate license negotiation'), copyright holders (UCSD, Stanford, UCSF), Data Access Committee oversight, citation requirements (bioRxiv article + data collection DOI). Governance: Data Access Committee supervision for ethical distribution and potential dual licensing

License choice (CC BY-NC-SA) semantically appropriate for academic research with commercialization potential. Copyright attribution to UC/Stanford/UCSF aligns with creator institutions. Data Access Committee governance matches NIH data sharing policy requirements.

Exceptional interoperability with comprehensive standards adoption across metadata (RO-Crate, Schema.org), biological ontologies (GO, Reactome, Cell Ontology), structural databases (PDB, AlphaFold), and domain repositories (MassIVE, SRA, NDEx). FAIRSCAPE framework ensures AI-readiness.

Standards documented: RO-Crate packaging, JSON-LD metadata with Schema.org vocabularies, EVI Evidence Graph Ontology, Gene Ontology (GO), Reactome pathways, Protein Data Bank (PDB), AlphaFold Database, NCI Thesaurus, BioAssay Ontology, Cell Ontology, CHEBI. Community repositories: MassIVE (proteomics), NCBI SRA (genomics), NDEx (networks). FAIRSCAPE framework for AI-readiness with JSON-Schema validation

Standard selections are domain-appropriate: GO/Reactome for functional annotation, PDB/AlphaFold for structural context, Schema.org/EVI for machine-readable metadata, MassIVE for proteomics community compliance. All standards actively maintained.

Excellent technical documentation covering full data processing pipeline from raw data acquisition through integration and annotation. Each strategy documented with specific methods and tools. Software tools identified with purposes.

preprocessing_strategies: 5 detailed strategies (deep learning embeddings, hierarchical community detection, cell map annotation, imaging QC, mass spec QC). cleaning_strategies: 3 strategies (FAIRSCAPE packaging, data standard mapping, integrative structure modeling). Software tools: node2vec (PPI embeddings), Human Protein Atlas deep learning model (image embeddings), Cytoscape (community detection), MuSIC pipeline (integration), FAIRSCAPE-CLI (validation), LLM annotation, Gene Ontology/Reactome alignment

Methodology is technically sound: node2vec appropriate for network embeddings, HPA model established for subcellular localization, Cytoscape standard for network analysis, FAIRSCAPE framework validated approach for AI-readiness. Processing workflow logically sequenced.

Comprehensive collection documentation with specific protocols, responsible laboratories, instrumentation details (confocal microscopy, 10x Genomics 3'HT kit), and clear release timeline. Each modality traced to expert laboratory.

collection_mechanisms: 4 detailed mechanisms (IF spatial proteomics, AP-MS, SEC-MS, CRISPR screens) with specific protocols. acquisition_methods: 4 methods (confocal microscopy, mass spectrometry, single-cell RNA-seq, MuSIC integration) with technical details. Data collectors: Lundberg Lab at Stanford (spatial proteomics), Krogan Lab at UCSF (PPI/SEC-MS), Mali Lab at UCSD (CRISPR). Timeframes: quarterly releases through Nov 2026, current releases v1.4 (March 2025), v2.1 (June 2025)

Laboratory assignments are accurate: Lundberg Lab known for spatial proteomics/Human Protein Atlas, Krogan Lab established in proteomics/PPI mapping, Mali Lab expertise in CRISPR screens. Timeline realistic for 4-year NIH project (2022-2026).

Excellent versioning infrastructure with version-specific DOIs, documented release contents, update timeline, errata tracking, and future roadmap. Each version archived in Dataverse for long-term access.

updates: detailed update plan with specific versions (alpha v0.5, beta v1.4 doi:10.18130/V3/B35XWX, beta v2.1 doi:10.18130/V3/F3TD5R), quarterly release schedule through Nov 2026, version-specific content documented (v1.4: perturb-seq, SEC-MS, IF images; v2.1: RGB IF images, metadata corrections, naming convention changes). version_access: via Dataverse with persistent DOIs. Errata handling: metadata corrections documented in v2.1. Future plans: computed cell maps and complete integration

Version progression logical: alpha supplemental → beta quarterly releases → final Nov 2026. DOI assignment per version follows best practices. Update frequency (quarterly) realistic for active data generation project.

Strong publication linkage with primary methodology paper, perturbation atlas paper, dataset DOIs, and federal grant documentation. Citation requirements built into license terms ensuring proper attribution.

external_resources: bioRxiv publication (Clark T, et al. Cell Maps for Artificial Intelligence, doi:10.1101/2024.05.21.589311), Perturbation Cell Atlas publication (Nourreddine S, et al., doi:10.1101/2024.11.03.621734, PMCID: PMC11580897), NIH RePORTER project details, dataset DOIs (10.18130/V3/DXWOS5, 10.18130/V3/B35XWX, 10.18130/V3/F3TD5R). license_and_use_terms: citation requirements for bioRxiv article and data collection DOI

bioRxiv DOI format correct (10.1101 prefix). PMCID format valid. Publication timeline realistic (May 2024, Nov 2024 preprints during active project 2022-2026). Citation requirements align with NIH data sharing policy.

Strong documentation of cell line origins including donor demographics (age, sex, race/ethnicity, disease status). Subpopulations clearly defined by treatment condition and differentiation state. Score 4/5 because this is not human subjects research - cell lines are de-identified and cannot represent current human diversity.

instances: detailed cell line characterization (MDA-MB-468: 51yo Black female, triple-negative breast cancer, metastatic site; KOLF2.1J: male Northern European donor, healthy). subpopulations: 7 experimental conditions documented (3 MDA-MB-468 treatments, 4 KOLF2.1J differentiation states). special_populations: donor demographics noted. Note: not human subjects research (de-identified commercial cell lines)

Cell line metadata is accurate: MDA-MB-468 from ATCC (RRID:CVCL_0419) matches published characterization, KOLF2.1J from HipSci (RRID:CVCL_B5P3) matches known iPSC line. Donor demographics align with cell line documentation.

Excellent findability with multiple persistent identifiers (DOIs), project website, federal grant records, and links to all deposition repositories and related platforms.

page: https://www.cm4ai.org, DOIs: 10.18130/V3/DXWOS5, 10.18130/V3/B35XWX, 10.18130/V3/F3TD5R, external_resources: 12 persistent URLs (cm4ai.org, NIH RePORTER, Dataverse DOIs, bioRxiv DOIs, fairscape.github.io, integrativemodeling.org, ndexbio.org, commonfund.nih.gov, nih-cfde.org, MassIVE, NCBI)

All URLs follow persistent identifier best practices. DOIs resolve to landing pages. Project website (cm4ai.org) active. Repository links (MassIVE, NCBI, NDEx, Dataverse) are authoritative platforms.

Excellent access documentation with multiple access pathways appropriate to each data type, clear licensing terms, and both human-friendly (web interfaces) and machine-readable (APIs, DOIs) access methods.

distribution_formats: 5 access mechanisms clearly documented (RO-Crate packages via Dataverse DOIs with machine/human-readable landing pages, Mass spec via MassIVE Repository, Sequences via NCBI SRA, Cell maps via NDEx with web browser/Cytoscape/HiView/Python ndex2 access). license_and_use_terms: clear terms (CC BY-NC-SA 4.0, non-commercial, attribution required, commercial requires separate license via Data Access Committee). Access paths: direct download from repositories, visualization tools (NDEx, Cytoscape), programmatic access (Python libraries)

Access mechanisms appropriate: Dataverse for archival access, MassIVE for proteomics community, NCBI SRA for genomics community, NDEx for network analysis. Tools mentioned (Cytoscape, HiView, ndex2) are established platforms.

Exceptional license clarity with explicit reuse permissions, attribution requirements, commercial restrictions, derivative work terms, and responsible parties for licensing negotiations. Addresses both open science and IP protection.

license: CC BY-NC-SA 4.0, license_and_use_terms: comprehensive reuse terms (attribution required to copyright holders and CM4AI project, cite bioRxiv article and data collection DOI, non-commercial use only, share-alike for derivatives, commercial use requires separate license from UCSD/Stanford/UCSF, Data Access Committee oversight). Copyright specified: (c) 2025 Regents of UC, (c) 2025 Stanford for spatial proteomics

CC BY-NC-SA 4.0 license appropriate for publicly funded research with commercialization potential. Attribution requirements comply with NIH data sharing policy. Copyright holders match project institutions. Share-alike terms ensure derivative openness.

Excellent provenance infrastructure using FAIRSCAPE framework for machine-readable provenance tracking. Version history documented with specific changes. Future updates planned. Provenance graphs trace data lineage from raw acquisition through MuSIC integration.

updates: detailed version history with timestamps (alpha v0.5, March 2025 v1.4, June 2025 v2.1), change documentation (v2.1: RGB IF images, metadata corrections, naming convention changes), future roadmap (computed cell maps, complete integration by Nov 2026). Provenance: FAIRSCAPE framework with machine-readable provenance graphs using EVI Evidence Graph Ontology, end-to-end provenance entailments, RO-Crate packaging with datasets/metadata/software, ARK persistent identifiers

FAIRSCAPE framework is validated approach for computational provenance (used in multiple NIH projects). EVI ontology appropriate for evidence graphs. ARK identifiers provide long-term persistence. Version numbering logical (0.5 alpha → 1.4 beta → 2.1 beta).

Excellent cross-platform interlinking with connections to federal grant records, multiple data repositories, publication databases, project websites, and biological databases. Metadata creates rich knowledge graph across platforms.

external_resources: 12 cross-platform links (cm4ai.org project portal, NIH RePORTER federal grants, UVA Dataverse archival repository, bioRxiv publications, FAIRSCAPE documentation, IMP structure modeling, NDEx network visualization, Bridge2AI program, NIH CFDE collaboration, MassIVE proteomics, NCBI genomics, Perturbation Atlas). Cross-references: dataset DOIs cited in publications, grant numbers link to federal records, cell line RRIDs link to Cellosaurus, standards link to GO/Reactome/PDB

Platform connections are semantically appropriate: Dataverse for long-term preservation, MassIVE for proteomics community, NCBI for genomics community, NDEx for network biology, bioRxiv for preprints, NIH RePORTER for grant transparency. All platforms actively maintained.