STAR RNA-seq Alignment - Usage Guide

Overview

STAR is the fast, splice-aware RNA-seq aligner that places reads across exon-exon junctions on the genome. It is the choice when the deliverable needs genomic coordinates -- novel isoforms, fusions, RNA variant calling, coverage tracks, splicing QC, or single-cell counting (STARsolo) -- and has memory to spare (~30 GB for human). The decisions that actually shape the result are the splice-junction database (built from a GTF at sjdbOverhang = readlength - 1), the per-sample vs cohort two-pass choice, the 255-for-unique MAPQ that breaks GATK, and the strand column STAR reports; this skill emphasizes those over the basic command.

Prerequisites

conda install -c bioconda star samtools

• A genome FASTA and a matching GTF annotation from the same provider/release (contig naming must agree; reconcile with alignment-files).
• ~30 GB RAM for a human index; route memory-constrained jobs to hisat2-alignment.
• The read length, to set --sjdbOverhang at index build.

Quick Start

Tell your AI agent what you want to do: - "Build a STAR index for my genome and GTF with the right sjdbOverhang for 150 bp reads" - "Align my paired-end RNA-seq reads with STAR in two-pass mode and get gene counts" - "Detect my library strandedness from the STAR GeneCounts output" - "Align RNA-seq for GATK variant calling without the empty-VCF MAPQ problem" - "Run STAR for fusion detection with STAR-Fusion"

Example Prompts

Index generation

"Create a STAR index for genome.fa with genes.gtf for 150 bp reads, and reduce genomeSAindexNbases appropriately because my genome is a 5 Mb bacterial assembly."

Two-pass and counts

"Align my paired-end RNA-seq with STAR two-pass, output a coordinate-sorted BAM and gene counts, and tell me from the counts whether my library is stranded."

RNA variant calling

"Align RNA-seq with STAR for GATK variant calling and make sure the unique-read MAPQ does not cause an empty VCF."

Cohort splicing

"I am doing differential splicing across 30 samples -- how do I run STAR two-pass so the junction set is identical across samples instead of per-sample?"

Fusion detection

"Run STAR with chimeric output for STAR-Fusion on my tumor RNA-seq."

What the Agent Will Do

1. Build the index with --sjdbGTFfile and --sjdbOverhang = readlength - 1, reducing --genomeSAindexNbases for small genomes.
2. Align to a coordinate-sorted BAM (STAR sorts natively; a later samtools sort is redundant), adding --quantMode GeneCounts to also read strandedness.
3. Choose per-sample two-pass for novel-junction work, or the cohort pooling recipe (shared SJ.out.tab) for splicing comparisons.
4. Set --outSAMmapqUnique 60 for any STAR -> GATK path so the 255 uniques are not dropped.
5. Avoid MAPQ-filtering RNA output for counting (it deletes multimappers), and route multimapper handling to the counter.
6. Reconcile genome/GTF contig naming and run the QC gate (alignment-files/bam-statistics) before counting or calling.

Index sjdbOverhang by Read Length

Strandedness from GeneCounts

Tips

• Set --sjdbOverhang to read length - 1 at index build; the default 100 quietly degrades junction sensitivity for short reads.
• STAR already coordinate-sorts with --outSAMtype BAM SortedByCoordinate -- do not run samtools sort again.
• For any STAR -> GATK RNA-variant path, set --outSAMmapqUnique 60 so the 255 uniques are not dropped as "unavailable," and inject read groups with --outSAMattrRGline ID:.. SM:.. PL:ILLUMINA LB:.. (space-separated tags, not bwa's -R '@RG\t...') because GATK requires them.
• STAR does not auto-detect gzip; pass --readFilesCommand zcat for .gz input (bwa/bowtie2/HISAT2 auto-detect it). For htseq-count, emit --outSAMtype BAM Unsorted or re-sort with samtools sort -n; featureCounts accepts coordinate order.
• Do not MAPQ-filter STAR output for counting; a -q 10 filter deletes every multimapper and under-counts gene families.
• For cohort splicing/sQTL, pool the pass-1 SJ.out.tab files and re-align all samples against one common junction set; per-sample two-pass is a batch effect.
• Infer strandedness from the GeneCounts columns rather than assuming; the wrong column roughly halves the counts.
• Reduce --genomeSAindexNbases for small genomes or STAR silently builds a broken index.

STAR vs HISAT2

Related Skills

• hisat2-alignment - Low-memory splice-aware alternative to STAR
• bwa-alignment - DNA short-read mapping (when reads do not cross junctions)
• read-qc/rnaseq-qc - RNA destination metrics: rRNA, gene-body coverage, strandedness
• read-qc/fastp-workflow - Trim adapters/poly-A before alignment
• alignment-files/bam-statistics - flagstat/idxstats QC gate; what a high mapping rate hides; contig naming
• rna-quantification/featurecounts-counting - Count aligned reads over genes (NH-aware multimapper handling)
• rna-quantification/alignment-free-quant - Salmon/kallisto when only known-transcript DE is needed
• differential-expression/deseq2-basics - Downstream DE from the count matrix
• single-cell/data-io - STARsolo single-cell counts into a single-cell workflow