HISAT2

HISAT2 aligns next-generation RNA sequencing (RNA-seq) reads to single and multiple reference genomes for spliced alignment and downstream gene-expression analysis.

Key Features:

• Efficient Indexing Scheme: Uses a hierarchical indexing approach based on the Burrows-Wheeler transform and the Ferragina-Manzini (FM) index with a whole-genome FM index plus numerous local FM indexes.
• Human-genome locality: The human genome hierarchical index comprises approximately 48,000 local FM indexes, each covering about 64,000 base pairs.
• Performance: Reported as a fastest RNA-seq aligner with equal or superior accuracy versus other methods and a reported memory footprint of 4.3 gigabytes.
• Scalability: Supports genomes of any size, including genomes exceeding 4 billion bases.
• Integration with analysis tools: Integrates with StringTie and Ballgown to enable read alignment, transcript assembly (including novel splice variants), computation of transcript abundance per sample, and comparison across experiments for differential expression.
• Benchmarking and optimization: Benchmarked against other splice-aware aligners using simulated data, showing superior base-, read-, and exon junction-level accuracy and demonstrating the importance of parameter optimization.

Scientific Applications:

• Transcript assembly and quantification: Alignments produced by HISAT2 feed transcript assemblers (e.g., StringTie) to assemble transcripts and quantify transcript abundance, including novel splice variants.
• Differential gene and transcript expression: Provides aligned reads for identification of differentially expressed genes and transcripts across samples and conditions.
• Comparative gene expression across species and conditions: Enables analysis of gene expression patterns by aligning RNA-seq reads to single or multiple reference genomes for cross-sample and cross-species studies.

Methodology:

Employs hierarchical indexing based on the Burrows-Wheeler transform and FM index with a whole-genome FM index plus numerous local FM indexes (~48,000 local indexes of ~64,000 bp in the human genome), performs splice-aware alignment, and has been benchmarked against other splice-aware aligners using simulated data to assess base-, read-, and exon junction-level accuracy with parameter optimization highlighted as important.