HISAT2 is a successor of TopHat.

TopHat has entered a low maintenance, low support stage as it is now largely superseded by HISAT2 which provides the same core functionality (i.e. spliced alignment of RNA-Seq reads), in a more accurate and much more efficient way.

Genome annotation file (GTF) is not required in the execution of HISAT2 itself.

First of all, index for HISAT2 should be constructed unless pre-calculated indexes are available (indexes for popular genomes are pre-calculated and can be downloaded from HISAT2 website ). When the FASTA-formatted genome is hogenome.fa and the name of that index is hoge, command to build the index is below.

hisat2-build -p 4 hogenome.fa hoge

-p option specifies the number of max processors to use (in this case, 4 is specified).

One example for shell script to run hisat2, named hisat2.sh, is like this.

hisat2 -p $p -x $x --dta -1 ${base}_1.fq.gz -2 ${base}_2.fq.gz \
| samtools view -@ $p -bS - \
| samtools sort -@ $p -T /tmp/hoge$$ -o ${base}.bam -

Because hisat2 outputs SAM-formatted alignment file by default, output is designed to be converted to BAM file using samtools view. Additionally the BAM file is designed to be converted to be sorted using samtools sort as the BAM file generated is not sorted, but the file is often needed to be sorted in other program.

Command to run the script is below.

time sh hisat2.sh sample

In this example, (gzipped) FASTQ files for paired-end reads are sample_1.fq.gz and sample_2.fq.gz.

It takes sometime to run hisat2, but the execution time for HISAT2 is much shorter than that for TopHat2.

Written by bonohu in rnaseq on 木 17 5月 2018.