REMBI Alignment: Integrating Biological Imaging Metadata Standards with BioStride

Executive Summary

REMBI (Recommended Metadata for Biological Images) represents a community-driven standard for biological imaging metadata that shares significant overlap with BioStride's goals for structural biology data management. Published in Nature Methods in 2021, REMBI provides guidelines for describing imaging experiments across light microscopy, electron microscopy, and X-ray microscopy—making it highly relevant to BioStride's multi-modal approach. This document analyzes the alignment between REMBI and BioStride, identifying opportunities for integration and mutual reinforcement of metadata standards in structural biology.

1. Introduction to REMBI

1.1 Background and Development

REMBI emerged from a critical need in the bioimaging community to standardize metadata for the vast amounts of imaging data being generated. Developed through a collaborative workshop in Hinxton (October 2019), REMBI brought together representatives from light, electron, and X-ray microscopy communities to establish common metadata guidelines.

Key Publication: Sarkans, U., Chiu, W., Collinson, L., et al. (2021). REMBI: Recommended Metadata for Biological Images—enabling reuse of microscopy data in biology. Nature Methods, 18, 1418–1422.

1.2 Core Objectives

REMBI aims to: - Enable systematic archiving of imaging data and metadata in public databases - Support FAIR (Findable, Accessible, Interoperable, Reusable) data principles - Facilitate automated data harvesting using machine learning techniques - Bridge diverse imaging communities with flexible yet comprehensive metadata standards

1.3 Adoption and Implementation

REMBI has been adopted by major imaging repositories: - BioImage Archive: Primary implementation platform - EMPIAR: Electron Microscopy Public Image Archive - Cell-IDR: Image Data Resource for cellular imaging - Tissue-IDR: Tissue-level imaging repository

2. REMBI's Eight-Component Structure

2.1 Hierarchical Organization

REMBI organizes metadata into eight high-level components that encompass diverse biological imaging methods:

REMBI_Structure:
  1_Study:
    description: "Top-level project metadata"
    alignment: "Dublin Core, DataCite, schema.org"

  2_Study_Component:
    description: "Container for organizing related data"
    example: "Separate components for EM and confocal in correlative studies"

  3_Biosample:
    description: "Biological sample metadata"
    importance: "Critical for biological context"

  4_Specimen:
    description: "Specimen preparation details"
    includes: "Fixation, staining, mounting"

  5_Image_Acquisition:
    description: "Microscopy technique and settings"
    audience: "Imaging scientists and microscopists"

  6_Image_Data:
    description: "Image file metadata"
    implementation: "File list with technical parameters"

  7_Image_Correlation:
    description: "Multi-modal image registration"
    status: "Optional"
    use_case: "Correlative microscopy"

  8_Analyzed_Data:
    description: "Image analysis and measurements"
    status: "Optional"
    format: "Typically tabular"

2.2 Target Audiences

REMBI addresses three primary user groups:

Imaging Scientists: Need detailed acquisition parameters, physical properties of instruments
Computer Vision Researchers: Require ground truth data, segmentations, labeled datasets
Biologists: Focus on biological context, experimental conditions, sample preparation

3. Alignment with BioStride

3.1 Structural Mapping

BioStride and REMBI share substantial conceptual overlap, with complementary strengths:

REMBI Component	BioStride Equivalent	Alignment Level	Notes
Study	Study	High	Direct conceptual match
Study Component	ExperimentRun	High	Both organize data collection sessions
Biosample	Sample	Very High	Core biological material description
Specimen	SamplePreparation	Very High	Preparation protocols and methods
Image Acquisition	Instrument + ExperimentRun	High	BioStride separates instrument from run
Image Data	Image (2D/3D) + DataFile	High	BioStride has specialized image classes
Image Correlation	Cross-modal relationships	Medium	BioStride handles via Study relationships
Analyzed Data	WorkflowRun + DataFile	High	Processing and derived data

3.2 Complementary Strengths

REMBI Strengths for BioStride

Comprehensive Imaging Focus: REMBI provides deeper metadata for microscopy-specific parameters
Community Consensus: Broad acceptance across imaging communities
Correlation Metadata: Explicit support for multi-modal image registration
Machine Learning Ready: Designed with AI/ML applications in mind

BioStride Strengths for REMBI

Structural Biology Depth: Specialized support for crystallography, SAXS/SANS, spectroscopy
Workflow Provenance: Detailed computational processing tracking
Multi-Scale Integration: From atomic to tissue-level organization
LinkML Foundation: Semantic web compatibility and validation

3.3 Integration Opportunities

Integration_Strategy:
  metadata_harmonization:
    - Map REMBI fields to BioStride schema
    - Extend BioStride with REMBI-specific attributes
    - Create bidirectional converters

  shared_vocabularies:
    - Adopt REMBI controlled terms for microscopy
    - Contribute structural biology terms to REMBI
    - Align with common ontologies (UBERON, CL, etc.)

  tool_ecosystem:
    - Support REMBI export from BioStride
    - Import REMBI-compliant data
    - Validation across both standards

4. Specific Relevance to Structural Biology

4.1 Cryo-EM Applications

REMBI's adoption by EMPIAR creates direct relevance for structural biology:

Case Study: EMPIAR-10061 - Dataset: β-galactosidase raw cryo-EM data (12.4 TB) - Original resolution: 2.2 Å - Reuse examples: - Reprocessed to higher resolution by multiple groups - Used for algorithm development - Training data for deep learning particle picking - Cloud processing pipeline development

This demonstrates how REMBI metadata enables extensive data reuse in structural biology.

4.2 Correlative Microscopy

REMBI's Image Correlation component is particularly valuable for multi-modal structural biology:

Correlative_Example:
  technique_1:
    type: "Cryo-EM"
    resolution: "3-5 Å"
    provides: "Molecular structure"

  technique_2:
    type: "Fluorescence"
    resolution: "200-300 nm"
    provides: "Cellular context"

  correlation_metadata:
    transformation_matrix: "3x4 affine"
    registration_error: "50 nm"
    fiducial_markers: "Gold particles"

  biostride_extension:
    add_rembi_correlation: true
    maintain_provenance: true
    cross_validate: true

4.3 Light Microscopy Integration

Many structural biology studies now incorporate light microscopy for context:

Live cell imaging before vitrification
Fluorescence-guided lamella preparation
CLEM (Correlative Light and Electron Microscopy)
Super-resolution microscopy for validation

REMBI provides the metadata framework for these techniques that BioStride can leverage.

5. Implementation Recommendations

5.1 Short-Term Integration (0-6 months)

Metadata Mapping

Create explicit mappings between REMBI and BioStride:

class REMBIBioStrideMapper:
    """Bidirectional mapper between REMBI and BioStride schemas"""

    def rembi_to_biostride(self, rembi_data):
        biostride_study = {
            'title': rembi_data['study']['title'],
            'description': rembi_data['study']['description'],
            'samples': self.map_biosamples(rembi_data['biosample']),
            'sample_preparations': self.map_specimen(rembi_data['specimen']),
            'instrument_runs': self.map_acquisition(rembi_data['image_acquisition']),
            'images': self.map_image_data(rembi_data['image_data'])
        }

        # Handle optional correlation data
        if 'image_correlation' in rembi_data:
            biostride_study['cross_modal_registration'] = \
                self.map_correlation(rembi_data['image_correlation'])

        return biostride_study

    def biostride_to_rembi(self, biostride_data):
        # Inverse mapping for REMBI compliance
        pass

Vocabulary Alignment

Harmonize controlled vocabularies:

Vocabulary_Mapping:
  sample_types:
    rembi: ["cell", "tissue", "organism", "molecule"]
    biostride: ["protein", "nucleic_acid", "complex", "cell", "tissue"]
    unified: ["protein", "nucleic_acid", "complex", "cell", "tissue", "organism", "small_molecule"]

  preparation_methods:
    rembi: ["fixation", "staining", "mounting"]
    biostride: ["vitrification", "crystallization", "purification"]
    unified: # Combine both sets with clear categories

  imaging_techniques:
    rembi: ["brightfield", "fluorescence", "confocal", "EM", "super-resolution"]
    biostride: ["cryo_em", "x_ray", "saxs", "sans", "ftir", "xrf"]
    unified: # Comprehensive technique ontology

5.2 Medium-Term Development (6-12 months)

Schema Extension

Extend BioStride to incorporate REMBI-specific fields:

# Addition to BioStride schema
ImageCorrelation:
  is_a: AttributeGroup
  description: "REMBI-compatible image correlation metadata"
  attributes:
    source_image:
      range: Image
      description: "Reference image for alignment"
    target_image:
      range: Image
      description: "Image to be aligned"
    transformation_type:
      range: string
      description: "Type of transformation (affine, deformable, etc.)"
    transformation_parameters:
      range: float
      multivalued: true
      description: "Transformation matrix or parameters"
    registration_error:
      range: float
      description: "Registration accuracy in nm"
    fiducial_markers:
      range: string
      description: "Type of fiducial markers used"
    correlation_coefficient:
      range: float
      description: "Correlation quality metric"

Validation Framework

Implement cross-standard validation:

class CrossStandardValidator:
    """Validate data against both REMBI and BioStride"""

    def validate(self, data):
        results = {
            'biostride_valid': self.validate_biostride(data),
            'rembi_valid': self.validate_rembi(data),
            'cross_compatible': self.check_compatibility(data),
            'warnings': [],
            'errors': []
        }

        # Check for conflicts between standards
        if results['biostride_valid'] and not results['rembi_valid']:
            results['warnings'].append(
                "Data valid for BioStride but missing REMBI requirements"
            )

        return results

5.3 Long-Term Vision (1-2 years)

Unified Imaging Metadata Standard

Work toward a unified standard that combines strengths:

Future_Standard:
  name: "Unified Biological Imaging Metadata (UBIM)"
  foundation: "LinkML"
  components:
    - structural_biology: "From BioStride"
    - light_microscopy: "From REMBI"
    - image_correlation: "From REMBI"
    - workflow_provenance: "From BioStride"
    - ai_readiness: "From both"

  governance:
    - committee: "Joint REMBI-BioStride working group"
    - stakeholders: ["wwPDB", "EMDB", "BioImage Archive", "IDR"]
    - update_cycle: "Annual"

6. Practical Examples

6.1 Example: Cryo-CLEM Study

Combining cryo-EM with fluorescence microscopy:

Study:
  title: "Mitochondrial ATP synthase in situ structure"

  # REMBI-style biosample
  biosample:
    organism: "Homo sapiens"
    cell_type: "HeLa"
    organelle: "Mitochondria"

  # REMBI-style specimen preparation
  specimen:
    fixation: "Vitrification"
    temperature: "-180°C"
    fluorescent_label: "MitoTracker Green"

  # BioStride-style experiment runs
  experiments:
    - technique: "fluorescence_microscopy"
      rembi_metadata:
        excitation_wavelength: "488 nm"
        emission_wavelength: "510 nm"
        objective: "100x/1.4 NA"

    - technique: "cryo_em"
      voltage: "300 kV"
      detector: "K3"
      pixel_size: "0.83 Å"

  # REMBI image correlation
  correlation:
    method: "Fiducial-based"
    markers: "100 nm gold"
    accuracy: "50 nm"
    software: "ec-CLEM"

6.2 Example: Multi-Scale Plant Imaging

From organ to molecular level:

Study:
  title: "Chloroplast structure under stress"

  study_components:
    - name: "Whole leaf imaging"
      technique: "optical_microscopy"
      rembi_compliance: true
      scale: "mm"

    - name: "Cellular imaging"
      technique: "confocal_microscopy"
      rembi_compliance: true
      scale: "μm"

    - name: "Chloroplast ultrastructure"
      technique: "cryo_em_tomography"
      biostride_compliance: true
      scale: "nm"

    - name: "Protein structure"
      technique: "single_particle_cryoem"
      biostride_compliance: true
      scale: "Å"

  cross_scale_correlation:
    registration_method: "Hierarchical"
    coordinate_system: "Unified 3D"
    metadata_standard: "REMBI+BioStride"

7. Benefits of Integration

7.1 For the Structural Biology Community

Broader Data Integration: Seamlessly combine molecular structures with cellular context
Enhanced Discoverability: REMBI-compliant data more findable in imaging repositories
Improved Reproducibility: Comprehensive metadata across scales
AI/ML Readiness: Standardized metadata for training algorithms

7.2 For the Imaging Community

Structural Context: Link cellular images to molecular structures
Advanced Analytics: Leverage structural biology computational tools
Multi-Modal Workflows: Streamlined correlative approaches
Validation Frameworks: Benefit from structural biology's rigorous validation

7.3 For Data Repositories

Interoperability: Exchange data between EMPIAR, BioImage Archive, and others
Reduced Redundancy: Shared metadata models reduce duplication
Unified Submission: Single pipeline for multi-modal studies
Enhanced Services: Offer cross-repository search and analysis

8. Challenges and Mitigation

8.1 Technical Challenges

Challenge	Mitigation Strategy
Schema complexity	Provide user-friendly tools and wizards
Vocabulary conflicts	Create mapping tables and conversion tools
Validation overhead	Implement automated validation pipelines
Storage requirements	Use hierarchical storage with metadata separation

8.2 Community Challenges

Challenge	Mitigation Strategy
Adoption resistance	Demonstrate clear benefits through case studies
Training needs	Develop comprehensive educational materials
Legacy data	Provide migration tools and services
Governance	Establish joint committees with clear mandates

9. Implementation Roadmap

Phase 1: Foundation (Months 1-3)

Establish working group with REMBI representatives
Create detailed field mapping documentation
Develop proof-of-concept converter tools
Identify pilot projects for testing

Phase 2: Integration (Months 4-9)

Implement bidirectional converters
Extend BioStride schema with REMBI fields
Develop validation frameworks
Launch pilot projects with early adopters

Phase 3: Deployment (Months 10-12)

Release integrated tools and documentation
Train repository staff and users
Process feedback and iterate
Plan long-term governance structure

Phase 4: Maturation (Year 2+)

Establish formal partnership with REMBI
Contribute to standard evolution
Expand adoption across communities
Develop advanced integration features

10. Conclusions

10.1 Strategic Importance

The alignment between REMBI and BioStride represents a critical opportunity to bridge the imaging and structural biology communities. With REMBI's broad adoption in biological imaging and BioStride's comprehensive structural biology support, integration of these standards will enable unprecedented multi-scale, multi-modal studies.

10.2 Mutual Benefits

Both standards benefit from integration: - REMBI gains access to structural biology's sophisticated data models and validation frameworks - BioStride gains established imaging metadata standards and community adoption - The scientific community gains a more complete picture from molecules to organisms

10.3 Path Forward

The path to integration is clear and achievable: 1. Start with metadata mapping and vocabulary alignment 2. Build practical tools for data conversion 3. Demonstrate value through pilot projects 4. Formalize collaboration through governance structures 5. Work toward eventual standard convergence

10.4 Call to Action

We recommend: 1. Immediate: Contact REMBI maintainers to discuss collaboration 2. Short-term: Implement REMBI support in BioStride tools 3. Medium-term: Joint workshops and training materials 4. Long-term: Co-develop next-generation unified standards

The convergence of REMBI and BioStride will accelerate scientific discovery by enabling researchers to seamlessly navigate from cellular contexts to atomic details, ultimately advancing our understanding of biological systems across all scales.

References

Sarkans, U., et al. (2021). REMBI: Recommended Metadata for Biological Images—enabling reuse of microscopy data in biology. Nature Methods, 18, 1418–1422.
REMBI Overview. European Bioinformatics Institute. https://www.ebi.ac.uk/bioimage-archive/rembi-help-overview/
Iudin, A., et al. (2016). EMPIAR: a public archive for raw electron microscopy image data. Nature Methods, 13, 387–388.
Williams, E., et al. (2017). The Image Data Resource: A Bioimage Data Integration and Publication Platform. Nature Methods, 14, 775–781.
Hartley, M., et al. (2022). The BioImage Archive – Building a Home for Life-Sciences Microscopy Data. Journal of Molecular Biology, 434(11), 167505.