Reference-gene stability
Ranks candidate reference genes by how stably they behave across your own samples, and says how many of them you actually need to normalise against.
What it does
Relative quantification is a ratio, and the denominator is a choice. Normalising against a single "housekeeping" gene assumes that gene is constant across your conditions — an assumption that is usually inherited rather than tested, and one that fails often enough to invert published results.
This tests it against your data. Every candidate is scored by how much its expression varies relative to every other candidate across your samples; the least stable is dropped, and the scoring repeats until a ranking remains. Genes that are genuinely constant vary together only because of the loading differences you are trying to remove, so they end up stable; a gene that responds to your treatment does not.
The second half answers a question the ranking does not: how many genes to use. Adding reference genes stops helping at some point, and the pairwise-variation measure shows where that point is for your samples rather than deferring to a convention.
When you'd use this
- Before committing to a normaliser, especially in a tissue, developmental stage or treatment where the usual candidates have not been validated.
- When a fold change is surprising. If the reference itself moved with the treatment, the effect you are seeing may belong to the denominator.
- When choosing how many references to run. Three genes cost three reactions per sample; this says whether the third is buying you anything.
- When reviewers ask why these reference genes. The ranking is the answer, computed on the samples in question.
What this computes
- A stability measure per candidate — the average pairwise variation of that gene against every other candidate, recomputed as the least stable are removed. Lower is more stable.
- The stepwise elimination order, so you can see which genes were dropped and in what sequence rather than only which survived.
- The pairwise variation between using n and n+1 reference genes, which is what says how many to include.
- The ranking itself, with the two most stable genes reported as a pair — the measure cannot separate the final two, and saying so is more honest than breaking the tie arbitrarily.
What it does not do
It measures stability within the samples you give it. A gene ranked stable across four conditions is not thereby stable in a fifth, and the ranking is only as good as the candidate set — it can only tell you which of the genes you tested is best behaved, not whether a better one exists.
Protein coverage
Draws each inferred protein's sequence with the identified peptides laid under the residues they cover, marking which peptides it shares with other proteins.
Digital PCR
Turns partition counts into absolute copies per reaction with a confidence interval, and shows the ratios between targets without needing a standard curve.