1. Define the analytical target

State the biological question, material and matrix, endpoint, reportable result and unit, expected range, required precision, decision context, throughput, and operating constraints. This target guides the plate and statistical design.

The worked target below turns a vague request for “a potency assay” into design inputs. Its values are illustrative, not universal requirements. Similar-looking curves used for absolute dose response, inhibition, interpolation, screening, or monitoring require different targets and evidence.

Illustrative analytical target profile
FieldWorked decision
QuestionPotency of test preparation relative to an assigned reference
EndpointIncreasing reporter response after 24 hours
ResultRelative potency in % with confidence interval
Intended range50% to 150% of reference
Routine designEight concentrations, triplicate wells, reference and test on one plate
Decision useSupport a defined lot-quality decision after suitability passes
ThroughputUp to four test preparations per analyst day

2. Draft the report and evidence path

Sketch the result, unit, uncertainty, source identity, well-level evidence, parameters, diagnostics, criteria, exclusion fields, decision, versions, and approvals a reviewer will need. Missing report fields expose missing experimental or calculation requirements while the design can still change.

Keep scientific values separate from display rules. Captions, digits, and styles may change without changing the retained calculation or its source-to-result identity.

3. Stabilize biology and plate execution

Select controls that represent the response window and known failure modes. Explore seeding, incubation, reagents, preparation, mixing, timing, plate type, instrument settings, and position effects.

Use designed experiments where practical instead of changing one factor until a curve looks acceptable. Record lots, instruments, operators, and timing so variation can be attributed rather than hidden in a final average.

4. Predefine the source-to-model data rules

Define source formats, measurement-region selection, plate roles, normalization, blank handling, dilutions, replicate use, exclusions, concentration units, transformations, and rounding before optimizing the model.

Retain raw and derived values together. Compare prespecified development alternatives openly, then approve one exact data path for routine execution instead of relying on analyst memory.

5. Evaluate the model and suitability as one method

Compare scientifically plausible models using representative runs and inspect designed range, residuals, parameter stability, uncertainty, known-sample performance, and repeated results—not only R².

Build suitability criteria from protected failure modes. Challenge calculations, missing values, boundary inclusivity, false passes, and false failures, then lock the selected model and criterion versions together.

6. Challenge, validate, and transfer the complete procedure

Robustness deliberately varies relevant conditions around target settings. Validation addresses the selected performance characteristics for intended use. Transfer or implementation verifies that the receiving environment can execute the controlled plate, calculation, review, and report as planned.

Treat each phase as a deliverable rather than an open-ended experiment. Version the method, reader profiles, layouts, software, analysis settings, criteria, report, training, and evidence before routine use.

PhaseMinimum retained deliverable
TargetApproved analytical target profile
Plate developmentSelected controls, range, layout, and operating settings with rationale
Data and modelExact transformation, fit, diagnostics, criteria, and expected-result fixtures
RobustnessDesigned challenges, results, affected parameters, and operating ranges
ValidationProtocol, complete execution evidence, deviations, and conclusion
TransferReceiving-site execution, training, comparison, gaps, and approval

Limits and product boundary

This sequence organizes scientific development; it does not prescribe one universal study design, model, threshold, or validation package. The accountable laboratory selects and validates the procedure for its intended use.

Provenarium can carry a supported relative-potency method—parallel-line log-log, common-shape 4PL, or common-shape 5PL—or a supported linear, 4PL, or 5PL standard-curve method from plate-reader data through controlled report evidence. Other models, endpoints, and source formats require a separate scope decision.

Frequently asked questions

When should assay software enter development?

Early enough to preserve data and compare analysis settings reproducibly, while keeping development work separate from approved routine execution.

Should system-suitability thresholds be copied from literature?

Literature can inform hypotheses, but thresholds should be justified for the assay, design, intended use, and observed performance.

Why record the analysis-software version?

It identifies the software that produced a result, is a prerequisite for an exact rerun when that release remains available, and supports assessment of later changes.

Primary references