Gene expression

Summary

Gene expression is a measured or inferred quantity in a defined biological and technical context. It is not a single fixed property permanently attached to a gene, and RNA abundance is not interchangeable with protein abundance or biological activity.

Core rules

• Expression can vary across tissues, cell types, developmental stages, physiological states, disease states, perturbations, and time.
• Gene-level and transcript-level expression are different quantities.
• RNA abundance, transcription rate, translation, protein abundance, and protein activity are related but distinct biological measurements.
• Bulk measurements can reflect changing cell composition as well as changes within cells.
• Counts, normalised abundance, TPM-like quantities, relative fold changes, and absolute molecule numbers are not interchangeable.
• Assay platform, library preparation, strandedness, annotation, quantification method, and normalisation can affect the reported value.
• Failure to detect expression is not equivalent to proving biological absence.
• A reference transcript selected for reporting does not establish that it is the predominantly expressed transcript in the biological system being studied.

Required context

Preserve as applicable:

• organism
• tissue, cell type, or cell composition
• developmental or disease state
• perturbation and time point
• assay type
• bulk or single-cell design
• gene-level or transcript-level quantification
• annotation provider and release
• normalisation method and unit
• comparison group and statistical model
• relevant batch, replicate, and study-design information

AI behaviour

• Qualify expression claims with biological system and measurement context when they affect interpretation.
• Do not infer absence of function from low or undetected expression in one unrelated tissue, cell type, condition, or assay.
• Do not treat bulk expression changes as cell-intrinsic without considering composition.
• Keep gene-level and transcript-level claims separate.
• Do not convert an assay-specific quantity into generic “expression level” when the unit or normalisation carries meaning.
• Do not infer protein abundance or activity directly from RNA abundance without supporting evidence.
• When comparing datasets, check annotation, quantification, normalisation, and biological context before attributing differences to biology.

Common failure modes

Context-free expression claim

“Gene X is not expressed” can be incorrect when the evidence only shows that it was not detected in one tissue, condition, assay, or dataset.

Bulk signal treated as cell-intrinsic

A change in bulk RNA abundance can arise because cell-type proportions changed rather than because each cell changed transcription.

RNA treated as protein activity

A transcript can be abundant while translation, localisation, degradation, post-translational modification, or inhibition limits protein function.

Transcript selection confused with expression

A MANE or canonical transcript can be selected for consistent reporting without being the most abundant transcript in every relevant tissue.

Authoritative standards

Use assay-specific reporting standards, annotation resources, and controlled vocabularies when exact metadata requirements matter. Use the relevant quantification and statistical method documentation for the meaning of units and normalisation.

Examples

Better statement

Prefer: “Gene X had low RNA abundance in this bulk liver RNA-seq dataset under the stated normalisation.”

Avoid: “Gene X has low expression” when the claim is intended to be universal.

Transcript-level statement

Prefer a statement that identifies the transcript, tissue, assay, and quantification method when transcript usage is central to the claim.

Sources

• GTEx Portal: https://gtexportal.org/
• ENCODE Project: https://www.encodeproject.org/
• Human Cell Atlas: https://www.humancellatlas.org/
• GENCODE: https://www.gencodegenes.org/