Skip to content

Understand the five processing stages

QC checks whether the raw reads are plausible before changing them. Use per-sample quality, adapter, duplication, and composition signals to decide whether the run is trustworthy.

fastp removes adapters and low-quality or too-short portions according to the selected cleanup profile. The run preserves preprocessing impact so you can see how many reads survived.

Salmon estimates transcript abundance efficiently. STAR + featureCounts aligns reads and counts features. The choice affects the count matrix used by DESeq2, so record it when comparing runs.

DESeq2 models count data across groups, estimates dispersion, and tests the configured contrasts. The baseline and design determine the sign and interpretation of log2 fold changes.

GSEA and pathway summaries translate ranked gene-level evidence into biological themes. They are downstream of differential-expression statistics and should be interpreted alongside QC and the selected contrast.