# Ontology Design

MCBO (Mammalian Cell Bioprocessing Ontology) uses a hub-and-spoke architecture with the 
Industrial Ontology Foundry (IOF) Core as the central hub, building on BFO foundations with 
domain-specific extensions.

## Hub-and-Spoke Architecture

MCBO reuses terms from multiple OBO Foundry ontologies:

| Ontology | Scope / Coverage | Example Terms Used | Role in MCBO |
|----------|------------------|-------------------|--------------|
| IOF Core | Industrial processes | ProductProductionProcess, hasOutput | Hub anchor for all processes/outputs |
| CLO | Cell lines | CHO-K1 | Standardized cell line IDs |
| CL | Cell types | Chinese hamster ovary cell | Biological cell type grounding |
| OBI | Assays | RNA-seq assay, has specified input/output | Experimental processes |
| EFO/ENVO | Conditions, environments | Hypoxia, culture pH | Capture experimental context |
| ChEBI | Chemicals | Glucose, L-glutamine | Media components |
| UO | Units | gram per liter | Normalize quantitative values |
| PATO | Phenotypes | Cell viability | Describe traits |
| SO/GO/PRO | Molecular entities | Transcript, gene ontology term, IgG protein | Link outputs to biology |

## Core Modeling Patterns

MCBO follows BFO-compliant semantic patterns to ensure interoperability with OBO Foundry ontologies.

### Process–Participant–Quality Chain

The core pattern for culture conditions:

```text
Process (BFO:process)
  └─ RO:0000057 (has participant) → CellCultureSystem (BFO:material entity)
      └─ RO:0000086 (has quality) → CultureConditionQuality (BFO:quality)
          ├─ hasTemperature (xsd:decimal)
          ├─ hasPH (xsd:decimal)
          └─ hasDissolvedOxygen (xsd:decimal)
```

**How it works:**

1. A bioprocess (e.g., `BatchCultureProcess`, `FedBatchCultureProcess`) is a `BFO:process` instance
2. The process `RO:0000057` (has participant) a `CellCultureSystem` (a `BFO:material entity`)
3. The `CellCultureSystem` `RO:0000086` (has quality) a `CultureConditionQuality` instance (a `BFO:quality`)
4. Temperature, pH, and dissolved oxygen values are attached as datatype properties to the `CultureConditionQuality` instance

This pattern preserves BFO semantics: processes have material entity participants, which bear qualities.
It avoids direct process-to-quality links that would conflict with BFO's occurrent–continuant distinction.

### Cell Line Engineering

```text
CellLine (material entity)
  └─ mcbo:overexpressesGene → Gene (information entity)
```

- Cell lines can `mcbo:overexpressesGene` gene individuals
- Inferred from Producer (boolean) + ProductType fields when explicit gene columns not present
- Antibody products (mAb/BsAb) use shared placeholder gene `mcbo:AntibodyProductGene`

### Sample Outputs

```text
CellCultureProcess (process)
  └─ mcbo:hasProcessOutput → BioprocessSample (material entity)
      └─ mcbo:inCulturePhase → CulturePhase (StationaryPhase, ExponentialPhase)
```

- Runs produce samples via `mcbo:hasProcessOutput`
- Samples can be in specific culture phases
- Productivity measurements are attached to runs

## Key Data Structures

| Structure | Location | Description |
|-----------|----------|-------------|
| TBox (Ontology) | `ontology/mcbo.owl.ttl` | Ontology schema - class definitions, properties |
| ABox (Instances) | `<data-dir>/mcbo-instances.ttl` | Instance data generated from CSV |
| Evaluation Graphs | `<data-dir>/graph.ttl` | Union of TBox + ABox for SPARQL queries |

## Scope and Boundaries

### In Scope

- Mammalian cell bioprocessing (CHO, HEK293)
- Culture process types: Batch, Fed-batch, Perfusion, Continuous
- RNA-seq data integration
- Culture conditions (temperature, pH, dissolved oxygen)
- Productivity measurements
- Cell line engineering (gene overexpression)

### Out of Scope

- Non-mammalian cell culture (yeast, bacterial)
- Downstream purification processes (future work)
- Proteomics data (future work)

## Dataset Classification

All dataset classes are subclasses of `IAO:dataset` (IAO_0000100), which is an information 
content entity (ICE) in the BFO/IAO hierarchy:

- `RNASeqDataset`
- `RawReadsDataset`
- `AlignedReadsDataset`

This ensures proper classification: datasets are ICEs, not material entities or processes.

## CSV to RDF Conversion

The `mcbo.csv_to_rdf` module transforms tabular metadata into RDF:

- Maps process types (Batch, FedBatch, Perfusion, etc.) to ontology classes
- Creates material entities (CellCultureSystem, cell lines, culture media)
- Attaches culture conditions (temperature, pH, dissolved oxygen) as qualities
- Handles productivity categorization (VeryHigh, High, Medium, LowMedium, Low)
- Infers gene overexpression from Producer + ProductType columns
- Generates IRI-safe identifiers from run/sample accessions

## SPARQL Query Architecture

Competency questions in `eval/queries/*.rq` leverage:

- `rdfs:subClassOf*` property paths for class hierarchies
- OBO relation IRIs (RO_0000057, RO_0000086) for standard relationships
- Filters on productivity types for optimization queries
- Cross-table relationship traversals via the RDF graph structure

## IOF Alignment

MCBO process classes (e.g., `CellCultureProcess`, `MammalianCellCultureProcess`) are direct 
subclasses of `BFO:process` to maintain strict BFO alignment and interoperability with 
OBO Foundry ontologies.

We use `rdfs:seeAlso` annotations to `iof:ManufacturingProcess` and 
`iof:ProductProductionProcess` to indicate conceptual alignment with IOF manufacturing 
concepts, while preserving direct BFO classification.

This approach ensures compatibility with both BFO-based biomedical ontologies and IOF-based 
industrial ontologies.