Quality Control

Quality control

All QC displays are on the Overview & QC tab and use only the included samples.

Display	What it shows	What to look for
Library size	Column totals	Samples far below the others
Detected genes	Genes with CPM ≥ 1 per sample	Low complexity or degraded samples
PCA	Top 2,000 variable genes, log₂ scale	Separation by group; outliers; batch structure
PC loadings	Genes driving each component	Whether a component reflects biology or artifact
MDS	limma plotMDS convention (leading log₂FC, top 500 genes per pair)	Same structure as PCA from a different distance
Sample dendrogram	Correlation distance, average linkage	Samples joining the wrong branch
Sample–sample correlation	Pearson correlation on log₂ values	A sample correlating better with another group
RLE	Relative log expression per sample	Boxes not centred on zero indicate a normalization problem
Expression density	Gaussian kernel density per sample (nrd0 bandwidth)	Curves that do not overlay
Library complexity and saturation	Detected genes against depth	Samples that would gain genes with more sequencing
Sex check (mouse, human)	Expression of sex-specific genes	Mismatch with recorded sex; sample swaps
Biotype composition	Share of expressed genes by biotype	Unexpected rRNA or mitochondrial content

Automatic warnings

The top of the tab lists flags such as low-depth samples and samples whose correlation profile resembles another group.

Excluding samples

Untick a sample in the sample bar to exclude it from every statistic. Exclusions are stored in session files and reported in the methods text. Record the reason for any exclusion.