# ADR 062: Specialize biomedical content for pathology

- HTML version: https://robbiepalmer.me/projects/personal-knowledge-graph/adrs/062-pathology-biomedical-content-specialization
- Project: Personal Knowledge Graph (https://robbiepalmer.me/projects/personal-knowledge-graph.md)
- Status: Accepted
- Date: 2026-11-18
- Initiatives: Semi-autonomous Software Development (https://robbiepalmer.me/initiatives/semi-autonomous-software-development.md)

## Context

[PDR 009](/projects/personal-knowledge-graph/pdrs/009-typed-biomedical-knowledge)
and [ADR 061](/projects/personal-knowledge-graph/adrs/061-typed-biomedical-content-nodes)
define the shared biomedical classes and routes. Pathology content still needs
rules for concepts that sit on different axes of laboratory work.

A tissue microarray describes how tissue cores are arranged. It does not state
how the donor tissue was preserved. A tissue microarray may therefore contain
FFPE cores and may later undergo H\&E staining, immunohistochemistry, or in situ
hybridization. Treating it as an alternative to FFPE would make valid
combinations impossible to represent. Published construction guidance likewise
describes cores taken from donor blocks and placed into a recipient block, then
analysed with several assays. [The ACSR construction and quality-assurance
guidance](https://pmc.ncbi.nlm.nih.gov/articles/PMC6813869/) is one example.

In situ hybridization is an assay family. Fluorescence and chromogenic
detection produce named specializations such as FISH and CISH. They are not
synonyms for every use of fluorescence or chromogenic detection. The
[NCI definition of FISH](https://www.cancer.gov/publications/dictionaries/cancer-terms/def/fluorescence-in-situ-hybridization)
describes fluorescently labelled DNA probes binding to targets in cells or
tissue.

The graph needs to preserve these distinctions without creating a separate
pathology graph or a new concrete node type for every laboratory variation.

## Decision

Keep pathology concepts in the concrete biomedical types accepted by ADR 061.
Do not add a generic pathology subject.

### Classify the initial pathology pages

Use these classes and boundaries:

| Biomedical class     | Initial pathology pages                                                        | Boundary                                                                            |
| -------------------- | ------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------- |
| Biomedical field     | Pathology, Histopathology, Molecular Pathology, Computational Pathology        | A field is not an assay, data type, or analysis technique.                          |
| Specimen preparation | FFPE tissue, Fresh-frozen tissue, Frozen section, Tissue microarray            | Preparations may describe preservation, sectioning, or arrangement and may coexist. |
| Assay method         | H\&E staining, Immunohistochemistry, Immunofluorescence, In situ hybridization | The assay remains distinct from its detection mode and output data.                 |
| Grading system       | Gleason grading, ISUP Grade Groups                                             | Architectural patterns, scores, and grade groups retain their stated meanings.      |

Multiple-instance learning and spatial analysis remain Idea nodes. Projects
may reference those ideas alongside pathology fields, preparations, assays,
data types, biomarkers, and grading systems.

A tissue microarray uses the `specimen-preparation` node type. Its page and
relations must state that it records specimen arrangement rather than
preservation. The graph may link the same project or future dataset to both a
tissue microarray and FFPE tissue. No exclusivity rule applies between
specimen-preparation nodes.

### Represent assay families with a typed edge

Add this relationship to the biomedical edge vocabulary:

| Edge           | Source       | Target       |
| -------------- | ------------ | ------------ |
| `SUBMETHOD_OF` | Assay method | Assay method |

Use `SUBMETHOD_OF` only for a narrower assay method whose definition depends
on the parent method. In situ hybridization is the parent method for later FISH
and CISH pages. A display label, acronym, detection mode, or implementation
detail does not create a subtype by itself.

The graph builder keeps forward and reverse indexes for `SUBMETHOD_OF` and
rejects cycles. Queries that return an assay family may include descendants
only when they request that traversal. A direct project edge continues to name
the most specific assay known from the project's evidence.

### Keep reader routes flat

Every page keeps the shared `/biomedicine/<slug>` route and Markdown twin.
Pathology remains both a landing-page facet and a field node. Page identities
omit a pathology prefix.
FISH and CISH can be deferred until project or data-product content needs
their pages without weakening the identity of the parent ISH page.

## Consequences

The graph can represent a study that uses an FFPE tissue microarray and applies
ISH without forcing those facts into one label. Readers can move from a
specialized assay to its parent while projects still record the most precise
method supported by their evidence.

The specimen-preparation class now contains concepts from more than one axis.
Consumers must not treat its members as mutually exclusive states. If later
content needs specimen formats outside laboratory preparation, a new concrete
class may replace this bounded treatment of tissue microarrays.

`SUBMETHOD_OF` adds cycle validation, reverse indexes, and a traversal choice
to assay queries. New assay pages must distinguish a real method specialization
from an acronym, detection mode, or implementation detail.

## Alternatives considered

### Add a tissue-array node type

This would separate arrangement from preservation at the type level. It was
rejected for the first tranche because tissue microarrays are the only required
array concept and the existing specimen-preparation type can represent the
page without an invalid edge. Revisit the class when another specimen format
needs the same relations.

### Model FISH and CISH as aliases of ISH

Aliases would preserve one page identity but lose the probes, detection,
evidence media, and project relations specific to each method. They remain
separate assay methods connected to ISH by `SUBMETHOD_OF`.

### Model fluorescence and chromogenic detection as assay methods

Detection alone does not identify the biological target or assay family. This
would recreate the ambiguity that typed biomedical nodes are meant to remove.
Keep detection details on assay content until a published use case justifies a
separate concrete class.

---

Markdown index of this site: https://robbiepalmer.me/llms.txt
