Before you start, identify the array platform and whether intensities have already been normalized and log2 transformed. RNA-seq counts and microarray intensities need different input routes.
1. Choose a source#
On Add data → Microarray, enter a GEO GSE accession or choose Import microarray matrix for a local table. An RNA-seq GEO series belongs on the public-read route and is redirected to SRA.
2. Set processing deliberately#
| Setting | What it changes | When to change it / example |
|---|---|---|
| GEO series matrix (default) | Uses the submitter-normalized matrix rather than reprocessing raw arrays. | Choose when the deposited processing matches your planned analysis; read the GEO processing description. |
| Affymetrix raw CEL → RMA | Reprocesses supported raw Affymetrix arrays with RMA. | Use when intentionally re-normalizing available CEL files; this is not a generic raw-format importer for every array platform. |
| Log2 transform: auto / force / off | Controls whether intensities are transformed. Double transformation distorts their scale. | Illustration: an intensity 256 becomes 8 after log2. If values are already log2 intensities, another log2 is not the same analysis. |
| Local matrix | Uses your processed intensities without a GEO download. | Confirm preprocessing with whoever supplied the matrix; a plausible-looking number range is not full provenance. |
3. Review sample groups and identifiers#
Fetched GEO arrays map probes to gene symbols; the platform accession is recorded. For a local gene matrix, ensure identifiers match the selected organism. Review condition labels and build the comparison on Samples and Analysis settings.
4. Run and review array-specific checks#
The route fits limma and writes a common differential-expression table schema. That filename/schema does not mean DESeq2 was fitted. Inspect normalization and probe-mapping checks before interpreting downstream figures or enrichment. There is no FASTQ, alignment or RNA-seq read-count QC for this route.
Advanced: design limits in the published baseline
Microarray limma supports the documented additive design route. Interaction and nesting operators are rejected in the published baseline; do not assume the same design capabilities as DESeq2. See advanced design.