bulk-rnaseq

Bulk RNA-seq — End-to-End Differential Expression Analysis Workflow

Skill Overview

The Bulk RNA-seq skill provides a complete, reproducible analysis workflow from raw FASTQ sequencing data to differentially expressed genes and enriched pathways, integrating quality control, alignment and quantification, count matrix construction, and statistical analysis.

Use Cases

From sequencing data to publication-ready results: Use this when you have raw FASTQ files or quantification outputs and need to run a complete, reproducible differential expression workflow—for example, “analyze my RNA-seq,” “FASTQ to DESeq2,” “run nf-core/rnaseq,” or “from sequencing data to differentially expressed genes and enriched pathways.”

Building a standardized count matrix: When you need to convert Salmon, STAR, or featureCounts outputs into a gene-level counts matrix and sample metadata template compatible with DESeq2/PyDESeq2, this skill provides bridging scripts and best-practice guidance.

RNA-seq experimental design and validation: Use this before large-scale analysis when you need to confirm whether the experimental design (number of replicates, batch effects, and strandedness) meets statistical requirements, or decide whether to use the nf-core standard workflow or an independent combination of tools.

Core Functions

Dual-path upstream processing: Supports both nf-core/rnaseq (Path A) and independent STAR/Salmon tools (Path B), two equivalent pathways that can be selected flexibly based on sample size, computing environment, and reproducibility requirements. Both ultimately generate a gene-level count matrix.

Counts matrix bridging: Uses build_counts_matrix.py to convert Salmon/STAR/featureCounts outputs into the integer counts matrix and metadata template required by PyDESeq2, handling key details such as gene ID mapping, selection of strandedness-specific columns, and rounding estimated counts to integers.

Complete workflow orchestration: Connects quality control (FastQC/MultiQC), alignment and quantification (STAR/Salmon), differential expression analysis (pydeseq2), pathway enrichment (pathway-enrichment), and visualization (scientific-visualization), emphasizing experimental design validation and QC gating to ensure defensible results.

Frequently Asked Questions

How many biological replicates are needed for RNA-seq analysis?

At least three biological replicates per group are required to achieve sufficient statistical power and stable dispersion estimates. Additional replicates are more important than deeper sequencing; with fewer than three replicates, it is almost impossible to obtain reliable differential expression results.

Should I use nf-core/rnaseq or perform a manual analysis?

Prefer nf-core/rnaseq (Path A)—it is a field-standard, audited, and citable workflow that integrates FastQC → trimming → STAR/Salmon → quantification → tximport → MultiQC in a single run. Manual analysis (Path B) is suitable for small numbers of samples, when you need to inspect each step in detail, or when running nonstandard steps unsupported by the workflow.

How can I determine RNA-seq strandedness?

Incorrect strandedness can silently result in the loss of approximately half of the reads. You can use Salmon’s -l A to infer it automatically or confirm it from the experimental design. After the run, verify the proportion of assigned reads and compare the different column outputs from STAR/featureCounts. This skill provides detailed determination methods in references/design-and-qc.md.