Statistical Reanalysis for a
Cell Signaling Experiment

RESEARCH QUESTION

What the Study Set Out to Test

This pseudoreplication example follows a cell signaling laboratory that silences a kinase with two independent siRNAs and measures a downstream protein by western blot. A reviewer has asked whether the statistics treat lanes as independent replicates, and the lab wants an independent statistical reanalysis before it responds.

The lab keeps authorship and decides what goes back to the reviewers. The need is a clear answer on the unit of analysis, a reanalysis that respects it, and a statement of how the conclusion changes, if at all. The primary question stays the same: does silencing the kinase change the abundance of the protein?

STUDY DESIGN AND DATA

Technical Lanes Within Independent Preparations

An experiment with three conditions (control siRNA and two kinase siRNAs), each with about six independent cell preparations, giving about 18 preparations. Each preparation runs on its own gel in three lanes, so there are about 54 lanes, and the original analysis counts all of them as replicates. The dataset holds band intensities, the loading control, preparation and gel identifiers and the condition.

METHOD AND ASSUMPTIONS

Why the Preparation Is the Unit

Lanes from one preparation share the same cells and the same gel, so they are repeated measurements of one biological replicate, not independent observations. Treating them as independent is pseudoreplication: it inflates the sample size from about 18 to about 54, so the standard errors and p-values become too small. The reanalysis fixes the unit before looking at results: the independent preparation is the biological replicate, and lanes are technical replicates within it.

A nested model, with lanes within preparations within conditions, separates the variation between preparations from the variation between lanes, and the comparison of conditions uses the between-preparation variation. The loading control is checked for stability under silencing before it is used, intensities are analyzed on the log scale, and the two primary comparisons, each kinase siRNA against control, are named in advance with a Dunnett adjustment for comparing two treatments with one control. Other proteins in the panel are labeled exploratory, and the reanalysis is described as a correction of the analysis, not a new experiment.

OUTPUT AND INTERPRETATION

What the Reanalysis Delivers

The output has the fold change for each kinase siRNA against control with a 95% confidence interval adjusted for the two comparisons, which is wider than the interval from the original analysis because the effective sample size is now the number of preparations. A figure shows every preparation as its own point beside the group summary, with the lanes visible within it, so a reader can see the data behind the estimate.

The write-up states plainly how the conclusion changes. If the intervals still exclude no change, the finding stands with its honest precision. If they are now compatible with no change, the memo says so and estimates how many preparations a follow-up experiment would need to settle the question. Agreement between the two siRNAs is reported as supporting an on-target effect, not as proof of it.

REPORTING

What the Researcher Received

A reanalysis memo the team can use with the editor and reviewers, with a methods paragraph that defines n and the unit of analysis, and a note for each issue the reviewer raised. The experimental interpretation, the response to reviewers and the decision on what to submit remain with the research team.