Citing Gradiance

Cite the software alongside the original dataset and the methods you used. A software citation identifies the tool; your methods section explains the analysis.

Software citation

Replace the bracketed fields with the version and access date from your analysis. Use a deployment or commit identifier when available; do not assume the current website version matches an older run.

Gradiance. (2026). Gradiance: Bioassay and flow cytometry analysis
(Version [version used]) [Software]. https://gradiancebio.com
Accessed [YYYY-MM-DD].

BibTeX

@software{gradiance,
  author  = {{Gradiance}},
  title   = {Gradiance: Bioassay and flow cytometry analysis},
  year    = {2026},
  version = {[version used]},
  url     = {https://gradiancebio.com},
  urldate = {[YYYY-MM-DD]}
}

RIS (EndNote, Mendeley)

TY  - COMP
AU  - Gradiance
TI  - Gradiance: Bioassay and flow cytometry analysis
PY  - 2026
ET  - [version used]
UR  - https://gradiancebio.com
Y2  - [YYYY-MM-DD]
ER  -

What to preserve

• Inputs: original files, source accession or DOI, license, units, sample identities, controls, and plate or channel mapping. Preserve a file checksum where possible.
• Experimental design: biological and technical replicate counts, grouping, paired measurements, and the experimental unit used for inference.
• Processing: blank subtraction, normalization, transformations, missing-value handling, and every exclusion with its reason.
• Analysis: model equation, parameter bounds, fitting method, weights, uncertainty method, statistical tests, and any multiple-testing correction.
• Execution: run identifier, date, software version, available code and dependency versions, and random seeds for stochastic methods.
• Review: the assistant instructions and relevant responses, decisions you changed, model diagnostics, and the final figures and numerical results.

Save the artifacts available in your account. PNG figures alone do not preserve the numerical results or processing steps. Code and additional export formats depend on availability and account entitlements; do not describe an export in your manuscript unless you obtained and checked it.

Understand what an export reproduces

For structured fit results, the Python export replays the saved observations, curves, parameters, and intervals. It writes the saved fit records to JSON and renders plots in compatible coordinate groups. It does not rerun the AI assistant, repeat preprocessing, or estimate the parameters again from the original input file. Recreating a figure from saved values verifies a different step from repeating the analysis.

To check a fit independently, preserve the original input and implement the reported model, preprocessing, bounds, weights, and uncertainty method in a recorded environment. Compare the resulting parameters and diagnostics with the saved results using a stated numerical tolerance. Report any analysis choices that are missing from the record instead of assuming them.

The audit bundle can preserve the available request, transcript, result, code trace, and input fingerprint. Inspect those fields: an older run may have a missing request, an empty code trace, or an unavailable input fingerprint after its temporary source file has been removed. A result or script hash helps detect changes to that content; a matching hash does not establish scientific validity, recover missing provenance, or verify that the script reruns the original analysis.

The Puromycin worked example separates a saved-result figure from an independent parameter check and records the limits of each.

Bioassay methods template

Use only statements supported by the completed run. Replace every bracketed field and remove clauses that do not apply.

Data from [source and assay] were analyzed in Gradiance [version; access date; run identifier]. Measurements were [preprocessing] and represented [n biological replicates and technical replicate structure]. We fitted [model equation] using [method, bounds, and weighting]. We report [parameter and units] with [uncertainty method and interval level]. Fit quality was assessed using [diagnostics]. [Exclusions and reasons, or no exclusions] were applied. The AI assistant was used for [tasks], and the authors reviewed [data mapping, analysis choices, and outputs]. Input data and available analysis artifacts are deposited at [repository and persistent identifier].

Flow cytometry methods template

FCS data from [source, instrument, panel, and acquisition conditions] were analyzed in Gradiance [version; access date; workspace identifier]. We applied [compensation or unmixing matrix and controls], [transform and parameters], and [quality-control exclusions]. Populations were defined by [gating hierarchy and controls], with frequencies reported relative to [parent population] and counts based on [events retained]. [Embedding or clustering method] used [channels, scaling, sampling, parameters, and seed]. The gating strategy, input metadata, and available workspace artifacts are deposited at [repository and persistent identifier].

Before submitting

Reopen the saved results and check that reported values match them. If you have executable code, rerun it against the preserved inputs in a recorded environment and compare numerical outputs within an explicit tolerance. A saved conversation or script is useful provenance, but does not by itself establish reproducibility.

Check your journal’s requirements for software and AI assistance disclosure. Gradiance is a research tool; authors remain responsible for the methods, interpretation, and claims in their work.

Continue with quality control, statistical methods, or the flow cytometry guide.