The problem. Everyone builds an RNA-seq workflow; far fewer can defend each choice. This review is the field’s consensus map — QC through interpretation — and I read it as an audit of my own plant_rnaseq_nf DAG rather than as new material.

The idea. It walks the whole pipeline and states the defensible default at each stage. Design first: biological replicates beat sequencing depth for detecting differential expression (three-plus replicates over more reads). QC at the read level (FastQC), then alignment — spliced aligners (STAR, HISAT) versus alignment-free quantification (kallisto, Salmon) and when each is appropriate. Quantification and normalization (why within- and between-sample normalization differ), differential expression (edgeR / DESeq2 / limma-voom and their assumptions), and functional interpretation (enrichment). Then the frontier: isoform-level analysis, fusion detection, eQTL, and integration with other omics.

Why it matters. Going stage by stage against my own pipeline is the most useful thing I can do with a review like this — it turns “I used STAR and DESeq2” into “here’s why, and here’s where the review says I could do better.” The replicates-over-depth point in particular is the kind of design wisdom that’s cheap to state and expensive to relearn.

Verdict. A few years on, the scaffold is still right even as specific tools moved (Salmon/kallisto now routine, single-cell exploded into its own literature). As a primary source it’s a review, so there’s no result to stress-test — but as a self-audit instrument it’s ideal. My plan: grade each node of my DAG against its recommendations in a follow-up post.