HISAT2 RNA-seq Alignment

Hard rules

1. No fabricated citations. Every cited work must resolve to a verifiable
2. No claim without provenance. Every quantitative or factual claim
3. No silent failure. Every script invocation, API call, or tool use must declare its exit status and what to do on non-zero. A skill that silently swallows errors is a violation.

When to use

• Bulk RNA-seq alignment to a reference transcriptome/genome.
• Small-genome RNA-seq where STAR's index cost is too high.
• 3' tag-seq (e.g., Lexogen QuantSeq) with appropriate flag settings.
• Long RNA-seq (>250 bp reads, e.g., ONT cDNA).

When NOT to use

• De novo transcript assembly → use STAR (better novel junction discovery).
• DNA-seq variant calling → use bwa-mem or bwa-mem2.
• Long reads → use minimap2 with -ax splice.

Prerequisites

• hisat2 ≥ 2.2
• samtools ≥ 1.19
• Reference FASTA + HISAT2 index
• For best sensitivity: known splice sites from a GTF file

Core workflow

1. Extract splice sites and exons from a GTF to build a sensitive index.
2. Build the HISAT2 index (hisat2-build or hisat2-build-s with known splice sites).
3. Align with hisat2 --dta for downstream transcriptome assembly (StringTie / Cufflinks) or --no-spliced-alignment for general use.
4. Sort and index the BAM with samtools.

Code patterns

Extract splice sites and exons from a GTF

hisat2_extract_splice_sites.py genes.gtf > splice_sites.txt
hisat2_extract_exons.py genes.gtf > exons.txt

Build a sensitive index (with known splice sites)

hisat2-build -p 16 --ss splice_sites.txt --exon exons.txt reference/genome.fa genome_hs2

Build a basic index (no annotation, faster)

hisat2-build -p 16 reference/genome.fa genome_hs2

Paired-end alignment (typical bulk RNA-seq)

hisat2 -p 16 --dta -x genome_hs2 \
    -1 reads_R1.fq.gz -2 reads_R2.fq.gz \
    --rg-id sample1 --rg SM:sample1 --rg PL:ILLUMINA --rg LB:lib1 \
    -S sample1.sam
samtools sort -@ 8 -o sample1.bam sample1.sam
samtools index sample1.bam
rm sample1.sam

--dta (downstream-transcriptome-assembly) is required if you'll run StringTie; it reports alignments tailored to transcript assembly.

Single-end alignment

hisat2 -p 16 --dta -x genome_hs2 -U reads.fq.gz --rg-id s1 --rg SM:s1 |
  samtools sort -@ 8 -o s1.bam -

rRNA-aware alignment (filter rRNA first, or use --un-gz to drop)

hisat2 -p 16 --dta --un-gz unmapped.fq.gz -x genome_hs2 -1 R1.fq -2 R2.fq |
  samtools sort -@ 8 -o s.bam -

--un-gz writes reads that didn't align (often rRNA or contaminant) for downstream QC.

Strand-specific libraries (dUTP / Ligation)

HISAT2 doesn't natively set XS tags; the convention is to use featureCounts or StringTie to infer strand from spliced alignments. For library prep-specific strand:

# Ligation protocol (Illumina TruSeq stranded)
# HISAT2 reports XS:A:+ / XS:A:- via the spliced alignment orientation
# featureCounts -s 2 (reverse stranded) will pick this up

Long-read cDNA (ONT / PacBio Iso-Seq)

hisat2 -p 16 --dta -x genome_hs2 -U iso_seq.fq.gz --no-temp-splicesite |
  samtools sort -@ 8 -o iso.bam -

nf-core integration

nf-core/rnaseq defaults to STAR. To use HISAT2, pass --aligner hisat2 on the CLI.

Common pitfalls

• Forgetting --dta for transcriptome assembly. StringTie/Cufflinks require the special scoring.
• Building an index without splice sites. The basic index works but is less sensitive for novel junctions.
• Mixing up --rna-strandness (legacy flag removed in 2.2). In HISAT2 2.2+, use downstream tools (featureCounts -s) to handle strand.
• Not using --rg-id. featureCounts and StringTie need read groups for multi-sample merges.
• STAR's memory cost is too high → HISAT2 with low sensitivity. If STAR fails on RAM, HISAT2 with the basic index will also miss junctions. Consider subsampling reads or using a known-splice-site index.

Validation

• samtools flagstat s.bam — high mapping rate expected (≥80% for human bulk RNA-seq).
• samtools view -c -f 2 s.bam — properly-paired count.
• samtools view s.bam | grep -c 'N:M:' — spliced alignment count (look for N in CIGAR).
• samtools view s.bam | awk '$6 ~ /N/' | wc -l — spliced reads.
• rRNA fraction should be < 10% for poly-A selected, < 25% for ribo-depleted.

Open alternatives

References

• HISAT2 paper: Kim et al. 2019, Nature Methods — 10.1038/s41587-019-0201-4
• HISAT2 manual: http://daehwankimlab.github.io/hisat2/manual/
• Companion: ors-bioinformatics-sequence-star-alignment, ors-bioinformatics-sequence-bwa-alignment.

Changelog

• 1.0.0 (2026-06-10): Initial adaptation by Pradyumna Jayaram from bio-hisat2-alignment (bioSkills-main/read-alignment/hisat2-alignment)."

Cross-references

Other skills in this category:

• batch-processing
• bowtie2-alignment
• bwa-alignment
• bwa-mem2-alignment
• codon-usage
• compressed-sequence-files
• fastq-quality-scores
• filter-sequences
• format-conversion
• motif-search
• paired-end-fastq
• pysam-genomics
• read-write-sequences
• reverse-complement
• sam-bam-basics
• samtools-bam-processing
• seq-objects
• sequence-properties
• sequence-slicing
• sequence-statistics
• star-alignment
• transcription-translation