Align RNA-seq reads with HISAT2

Build a HISAT2 splice-aware index and align paired RNA-seq reads to a genome, piping straight into samtools for a sorted BAM.

SSSudipta SardarUpdated July 21, 2026

Prerequisites

HISAT2 installed (`hisat2`, `hisat2-build`)
samtools installed
A reference genome FASTA file
Paired-end RNA-seq FASTQ files (R1 and R2)

1Build the HISAT2 index

Index your genome FASTA once per reference. This writes eight binary files (ref_index.1.ht2 through ref_index.8.ht2) that every later alignment reuses, so you never rebuild it for subsequent samples.

bash

hisat2-build reference.fasta ref_index

Expected output

Settings:
  Output files: "ref_index.*.ht2"
  Line rate: 6 (line is 64 bytes)
  Lines per side: 1 (side is 64 bytes)
  Offset rate: 4 (one in 16)
  FTable chars: 10
  Strings: unpacked
  Local offset rate: 3 (one in 8)
  Local fTable chars: 6
  Endianness: little
  Actual local endianness: little
  Sanity checking: disabled
  Assertions: disabled
  Random seed: 0
Reading reference sizes
  Time reading reference sizes: 00:00:06
Calculating joined length
Writing header
Reserving space for joined string
Joining reference sequences
  Time to join reference sequences: 00:00:03
Total time for constructing forward index: 00:02:28
Total time for call to driver() for forward index: 00:03:12

2Align and sort into a BAM

Align both mates against the index and stream the output straight into samtools sort, so no intermediate SAM ever touches disk. For a stranded dUTP/Illumina library, add the strandness flag before -x so read orientation is recorded: `hisat2 -p 4 --rna-strandness RF -x ref_index -1 R1.fastq.gz -2 R2.fastq.gz | samtools sort -@ 4 -o rna.sorted.bam -`.

bash

hisat2 -p 4 -x ref_index -1 R1.fastq.gz -2 R2.fastq.gz | samtools sort -@ 4 -o rna.sorted.bam -

Expected output

24500000 reads; of these:
  24500000 (100.00%) were paired; of these:
    2200000 (8.98%) aligned concordantly 0 times
    19800000 (80.82%) aligned concordantly exactly 1 time
    2500000 (10.20%) aligned concordantly >1 times
    ----
    2200000 pairs aligned concordantly 0 times; of these:
      500000 (22.73%) aligned discordantly 1 time
    ----
    1700000 pairs aligned 0 times concordantly or discordantly; of these:
      3400000 mates make up the pairs; of these:
        2783200 (81.86%) aligned 0 times
        400000 (11.76%) aligned exactly 1 time
        216800 (6.38%) aligned >1 times
94.32% overall alignment rate

Troubleshooting

Low overall alignment rate (well under 70%) on eukaryotic RNA-seq

Confirm you are aligning with HISAT2 (splice-aware), not a DNA aligner like BWA, and that the index was built from the genome FASTA, not a transcriptome. Reads spanning exon-exon junctions only map when both the aligner and the index are splice-aware.

Strand information looks wrong in downstream featureCounts/htseq counts

Set `--rna-strandness RF` for dUTP/Illumina stranded libraries, or `FR` for the opposite orientation. The default is unstranded, which mislabels strand for stranded protocols and skews per-gene counts.

References

HISAT2 manual