eduKateSG · Why Science?
Turn a screening result into an assay you can understand, reproduce and compare
Follow the assay from biological question through format, method, screened entity, endpoint and analysis before deciding whether two results belong together.
A screening table can contain thousands of numbers and still be hard to reuse if nobody can tell what each assay actually did. The BioAssay Ontology, or BAO, establishes shared metadata terms and definitions for low- and high-throughput drug and probe screening assays and their results. Its official repository centres concepts such as assay biology, method, format, endpoint and screened entity so data can be integrated, retrieved and analysed meaningfully. A 2026 peer-reviewed release paper describes version 2.8.16 extensions for pharmacokinetics, pharmacodynamics and safety pharmacology, together with improved alignment to resources including Uberon and ChEBI. BAO organises assay knowledge, but a structured endpoint does not prove biological mechanism, clinical effectiveness, safety or suitability for any person. This guide connects BAO to Secondary Science, O-Level Biology, Chemistry, mathematics, computing and STEM education; it is educational rather than laboratory, medical, regulatory, admission or career advice.
Related eduKate reading: Ontology for Biomedical Investigations; ChEBI stable chemical meaning; Protein Ontology; Statistics Ontology; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 13 October 2026: Official BioAssay Ontology repository, checked 13 October 2026; Official BioAssay Ontology project site; 2026 BAO release paper indexed by PubMed; Peer-reviewed BAO modularisation and applications paper; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Ontologies, definitions and release products can change, so preserve dated documentation, exact source records and file versions.
How a bioassay becomes comparable, reviewable evidence
A bioassay is more than a result column. Its biological system, perturbation, method, format, detection technology, endpoint and analysis pipeline jointly determine what a number means.
The official BAO project says its purpose is to supply common reference metadata for low- and high-throughput drug and probe screening. This enables integration only when source detail is retained.
Assay biology identifies the biological process, target, pathway, phenotype or model under investigation. Similar labels can hide different species, cell lines, constructs or disease models.
Assay format describes the experimental arrangement, such as biochemical, cell-based, tissue or organism-level contexts. Format changes the mechanisms and confounders that are plausible.
Assay method and detection technology describe how the response becomes a signal. Fluorescence, luminescence, imaging or label-free methods carry different artefacts and calibration needs.
The screened entity may be a chemical, biologic, genetic perturbation or other tested material. Stable identity needs batch, purity, concentration, formulation and provenance where relevant.
An endpoint is a defined summary of response, not the raw observation itself. Activity, potency, efficacy, toxicity and pharmacokinetic measures answer different questions.
The 2026 BAO paper reports extensions for pharmacokinetics, pharmacodynamics, safety pharmacology, disease models and toxicological measurements. Current release details therefore matter.
Ontology reasoning can classify assays through their components, but it cannot rescue poor controls, plate effects, contaminated cultures or an invalid statistical model.
The happy scientific promise is reusable context: a result can travel beyond its original spreadsheet without losing the design that gives it meaning.
Begin with the assay objective. Target engagement, pathway modulation, cell viability and organism behaviour can all produce numerical responses while supporting different claims.
Name positive, negative, vehicle and blank controls and explain what each control detects. A control class is not enough without its expected behaviour.
Record biological material precisely. Species, cell line, passage, genotype, tissue, construct and culture conditions can change the assay response.
Record the screened entity at the tested form. Salt, stereochemistry, purity, lot and storage may explain differences between nominally identical compounds.
Keep concentration and time series. A single high-dose observation cannot stand in for a full dose-response or a durable effect.
Preserve raw signal, normalised response and fitted parameter. Each transformation introduces assumptions that reviewers should be able to inspect.
Document curve models, bounds and failure codes. A reported potency without fit diagnostics may be less informative than the complete response series.
Separate technical repeats, biological replicates and independent experiments. Their variance estimates answer different reliability questions.
For classroom work, enzyme or photosynthesis investigations can be annotated by biology, format, method, endpoint and control without handling unsafe materials.
Reader-ready assay summaries pair BAO terms with a concise methods sentence and a direct route to source measurements.
Comparability is earned component by component. Two assays that share a target may use different formats, constructs, readouts, temperatures or analysis windows and should not be pooled automatically.
Assay interference can mimic activity. Autofluorescence, aggregation, quenching, cytotoxicity and instrument saturation need counter-screens or orthogonal evidence.
Hit selection is a rule, not a natural boundary. Preserve thresholds, replicate policy and quality metrics so another team can reproduce the selected set.
Plate maps and run order are evidence. Edge effects, drift and batch structure can produce patterns that resemble biology.
Pharmacokinetic and pharmacodynamic annotations require explicit matrices, species, routes, compartments, time points and derived parameter methods.
Safety pharmacology terms support structured study description, not a declaration that a compound is safe. Clinical and regulatory conclusions require broader evidence and authorised review.
Machine-learning datasets should retain assay provenance. Random splits can leak near-duplicate compounds, series or assay batches and exaggerate performance.
Release migration should compare class definitions, imports and inference results and report which assay groups or endpoints moved.
A reproducible BAO bundle contains protocols, plate layouts, material identities, raw signals, normalisation, curve fits, ontology releases, mappings, quality flags and code.
Students can learn that a graph point is the end of a chain from question to design, control, measurement and calculation—not an isolated fact.
Science tuition can use a familiar rate experiment to practise fair testing, tables, graphs, outliers and evidence-proportional conclusions.
Before release, benchmark the pipeline with valid assays, interference cases, failed fits, missing controls, unit mismatches and ontology migration examples.
A final audit asks whether every result can be reconstructed from raw observations and whether the chosen endpoint truly answers the stated biological question.
A strong BAO case study begins with two assays that both produce a number called “percent activity”. In the first, a purified enzyme and a fluorescent substrate are measured after a short incubation. In the second, living cells are imaged after a longer exposure and the reported number is derived from viable-cell counts. Students identify the biology, format, screened entity, concentration, method, detection technology, endpoint and controls for each assay. They soon see that identical column labels do not make the measurements equivalent. The first result may reflect enzyme inhibition plus optical interference; the second may reflect target biology, transport, metabolism, general cytotoxicity and image-analysis choices. BAO can describe these components so software retrieves appropriate comparisons, while the protocol and raw evidence explain what the result can mean. A production validation pack should go further. Include biochemical and cell-based assays, single-point and dose-response designs, valid and failed curve fits, multiple endpoint directions, mixed units, missing controls, signal saturation, compound autofluorescence, plate-edge effects and replicated runs. Preserve the plate map and run order so analysts can distinguish biological pattern from batch structure. Require every derived endpoint to name its calculation, transformation, fit model, parameter constraints and failure code. Quality rules should never convert a failed fit into zero activity or discard an outlier without an auditable reason. For pharmacokinetic and pharmacodynamic extensions, keep species, biological matrix, route, dose, compartment, sampling times and estimation method; a clearance or exposure parameter cannot be compared safely without them. Safety-pharmacology annotations require an equally careful boundary: structured assay evidence can support review, but it cannot declare a medicine safe or suitable for an individual. Machine-learning use introduces new tests. Split by chemical series, assay family or time where appropriate; check near duplicates across training and evaluation; and report performance per assay class rather than only one average. When the BAO release changes, rerun benchmark queries and show how classifications, feature sets and result counts moved. A reader-ready output should let a scientist follow any hit back through ontology term, protocol, controls, raw signal and calculation. For students, the same chain teaches the heart of experimental science: a result is trustworthy not because the spreadsheet is large, but because the question, fair comparison, measurement and reasoning remain visible.
Before treating two screening results as comparable, run a component-by-component concordance review. Confirm target or phenotype, biological material, format, method, detection technology, exposure schedule, control design, endpoint formula, unit, analysis population and quality rule. Mark each dimension as equivalent, transformable, materially different or unknown. Only transformable fields should be normalised, and every conversion should preserve the original value. If a result depends on censoring, curve constraints or a threshold, publish that rule beside the parameter. Then perform an orthogonal challenge: ask whether a different measurement technology supports the same biological interpretation. Agreement can strengthen confidence; disagreement can reveal interference, model assumptions or context dependence. For an educational graph, label axes, units, controls and uncertainty and show at least one failed or ambiguous case rather than presenting only a perfect curve. This teaches learners that quality control is part of discovery, not an embarrassing step to hide. The final release note should state the BAO and import versions, the validation set, changed classifications and unresolved mappings. That modest discipline turns a screening archive from a pile of attractive numbers into scientific evidence another person can interrogate and rebuild.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Design | What biological question does the assay ask? | Target, system, format, perturbation and controls | A defined assay design |
| Measure | How is response detected and summarised? | Method, instrument, endpoint, unit and time | A traceable measurement |
| Analyse | How is raw signal converted into a result? | Normalisation, curve fit, thresholds and uncertainty | A reproducible derived result |
| Compare | Which assays can be combined? | BAO mappings, versions, quality and domain review | A bounded comparison |
Inside this guide
1–12 · Biology, format and method
- 1. Begin with the biological question
- 2. Use the BAO identifier
- 3. Read the definition
- 4. Describe assay biology
- 5. Describe assay format
- 6. Describe assay method
- 7. Describe detection technology
- 8. Identify the screened entity
- 9. Identify the biological target
- 10. Identify the endpoint
- 11. Preserve raw signal
- 12. Preserve controls
13–24 · Endpoints, results and validation
25–36 · Learning, interpretation and pathways
- 25. Challenge batch effects
- 26. Challenge model leakage
- 27. Review BAO migrations
- 28. Connect to Primary Science
- 29. Connect to Secondary Science
- 30. Connect to mathematics
- 31. Connect to computing
- 32. Connect to pathways
- 33. Did you know?
- 34. Write a reader-ready assay note
- 35. Build a validation ladder
- 36. Finish with a reproducible BAO bundle
Section 1 of 36
1. Begin with the biological question
A screen is meaningful only when the desired biological evidence is explicit.
Working checkpoint — BAO identifier: Write the mechanism or phenotype being tested. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 2 of 36
2. Use the BAO identifier
Stable classes connect local assay descriptions to maintained definitions.
Working checkpoint — assay biology: Store identifier, label, definition and release. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 3 of 36
3. Read the definition
Similar assay labels can hide differences in system, method or endpoint.
Working checkpoint — assay format: Compare formal scope with the protocol. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 4 of 36
4. Describe assay biology
Target, pathway, phenotype and disease model define the biological context.
Working checkpoint — assay method: Name the level actually interrogated. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 5 of 36
5. Describe assay format
Biochemical, cell-based, tissue and organism formats introduce different constraints.
Working checkpoint — screened entity: Record the physical arrangement. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 6 of 36
6. Describe assay method
The procedural strategy explains how the system is perturbed and observed.
Working checkpoint — assay endpoint: Link to a versioned protocol. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 7 of 36
7. Describe detection technology
Instrument and signal technology shape noise, sensitivity and artefacts.
Working checkpoint — BAO identifier: Store settings and calibration. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 8 of 36
8. Identify the screened entity
Tested chemicals, proteins or genetic perturbations need stable identity and provenance.
Working checkpoint — assay biology: Record exact form and batch. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 9 of 36
9. Identify the biological target
Target identity may involve a protein form, complex, organism or cellular process.
Working checkpoint — assay format: Use the narrowest supported concept. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 10 of 36
10. Identify the endpoint
An endpoint summarises a response under a declared calculation.
Working checkpoint — assay method: Keep formula, unit and direction. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 11 of 36
11. Preserve raw signal
Primary readings allow alternative normalisation and quality review.
Working checkpoint — screened entity: Do not retain only the fitted number. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 12 of 36
12. Preserve controls
Controls establish baseline, expected response and technical validity.
Working checkpoint — assay endpoint: Name role and acceptance range. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 13 of 36
13. Represent dose
Concentration, unit, dilution series and actual exposure affect interpretation.
Working checkpoint — BAO identifier: Validate conversions. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 14 of 36
14. Represent time
Incubation, exposure, readout and sampling times may differ.
Working checkpoint — assay biology: Use explicit intervals. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 15 of 36
15. Represent replication
Technical and biological replicates estimate different sources of variation.
Working checkpoint — assay format: Keep replicate hierarchy. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 16 of 36
16. Connect ChEBI
Chemical ontology terms stabilise molecular identity across screens.
Working checkpoint — assay method: Preserve tested form and concentration. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 17 of 36
17. Connect Protein Ontology
Protein forms and modifications can distinguish biologically different targets.
Working checkpoint — screened entity: Store construct and isoform evidence. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 18 of 36
18. Connect Uberon and Cell Ontology
Anatomy and cell types clarify model systems while culture state still matters.
Working checkpoint — assay endpoint: Record source and preparation. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 19 of 36
19. Connect OBI
OBI describes investigation design, roles, specimens, devices and processes around BAO.
Working checkpoint — BAO identifier: Link assay to the wider study. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 20 of 36
20. Connect STATO
Statistical methods and derived endpoints need named models and assumptions.
Working checkpoint — assay biology: Keep analysis provenance. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 21 of 36
21. Build a result chain
Raw reading, normalised response, curve fit and hit call should remain linked.
Working checkpoint — assay format: Store every transformation. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 22 of 36
22. Challenge target-only grouping
Shared targets do not make assay format, method or endpoint equivalent.
Working checkpoint — assay method: Compare all components. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 23 of 36
23. Challenge endpoint names
IC50, EC50, percent activity and viability have distinct definitions and directions.
Working checkpoint — screened entity: Verify formula and units. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 24 of 36
24. Challenge interference
Signal artefacts can produce apparent activity without the proposed biology.
Working checkpoint — assay endpoint: Use counter-screens. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 25 of 36
25. Challenge batch effects
Plates, runs, operators and reagent lots can structure results.
Working checkpoint — BAO identifier: Inspect randomisation and controls. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 26 of 36
26. Challenge model leakage
Related compounds or duplicated assays across splits can inflate prediction performance.
Working checkpoint — assay biology: Use grouped external validation. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 27 of 36
27. Review BAO migrations
The 2026 release expanded PK, PD and safety-pharmacology coverage and related mappings.
Working checkpoint — assay format: Run an impact report. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 28 of 36
28. Connect to Primary Science
Fair tests and controls show why one changed variable needs careful measurement.
Working checkpoint — assay method: Use safe household examples. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 29 of 36
29. Connect to Secondary Science
Enzymes, cells, rates and dose response connect biology and chemistry to evidence.
Working checkpoint — screened entity: Annotate one syllabus investigation. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 30 of 36
30. Connect to mathematics
Normalisation, curves, uncertainty and thresholds reveal how endpoints are calculated.
Working checkpoint — assay endpoint: Recreate a small result table. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 31 of 36
31. Connect to computing
Ontologies, plate schemas, pipelines and quality rules make screening data reusable.
Working checkpoint — BAO identifier: Test failed and edge cases. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Section 32 of 36
32. Connect to pathways
BAO skills appear in pharmacology, assay development, bioinformatics, data science and laboratory informatics.
Working checkpoint — assay biology: Check current official routes separately. Preserve raw signal, transformations, controls, fit model and uncertainty.
For computing, make failed fits and inconsistent units visible.
Section 33 of 36
33. Did you know?
A 2026 BAO paper reports major additions for pharmacokinetics, pharmacodynamics and safety pharmacology in release 2.8.16.
Working checkpoint — assay format: Ask which release a dataset used. Separate assay classification from experimental quality and interpretation.
For writing, state the assay conditions before the result.
Section 34 of 36
34. Write a reader-ready assay note
Name biology, format, method, entity, controls, endpoint, result, release and limitations.
Working checkpoint — assay method: Link to raw data. Test batch effects, interference, unit conversions and ontology migrations.
For students, ask what the assay measures directly.
Section 35 of 36
35. Build a validation ladder
Check identity, protocol, controls, signal, endpoint, statistics, artefacts, ontology and review.
Working checkpoint — screened entity: Stop at the earliest failed layer. Do not turn screening activity into a clinical or safety claim.
For parents, the calm checkpoint is whether the learner can explain the control.
Section 36 of 36
36. Finish with a reproducible BAO bundle
Deliver protocols, materials, raw data, analysis, BAO files, mappings, tests and limitations.
Working checkpoint — assay endpoint: Make every hit rebuildable. Keep biology, format, method, screened entity, endpoint and result together.
For mathematics, trace the raw signal into the reported parameter.
Contents · Previous section · Continue to the Science Learning Hub