mrsFAST-Ultra

mrsFAST-Ultra performs cache-oblivious, SNP-aware short-read mapping to reference genomes for sensitive and scalable multi-mapping alignment of high-throughput sequencing data.

Key Features:

• Cache-Oblivious Design: Uses a cache-oblivious algorithmic design to optimize memory access patterns and reduce CPU operations per alignment across diverse hardware configurations.
• Compact Index Structures: Employs novel compact index structures that reduce memory footprint and computational demands, producing an index approximately 10 times smaller than mrsFAST.
• Multi-Mapping Efficiency: Reports multiple mappings per read and can return reads with up to a specified number of mapping loci within an error threshold instead of only the single best location.
• SNP-Aware Mapping: Discounts mismatches at common single nucleotide polymorphism (SNP) locations provided by dbSNP to increase the number of accurately mappable reads.
• Performance and Scalability: Parallelized across multiple cores and processors and tunable for different memory settings, yielding approximately fivefold speed improvement over mrsFAST and higher sensitivity and speed than Bowtie2 in multi-mapping modes.
• Index Size Efficiency: Produces an index of about 2 GB for the entire human reference genome, roughly half the size required by Bowtie2.

Scientific Applications:

• High-throughput sequencing analysis: Enables sensitive short-read mapping and comprehensive multi-mapping support for large-scale sequencing datasets.
• Structural variation detection: Improves read mappability and sensitivity for analyses that rely on accurate multi-mapping alignments.

Methodology:

Cache-oblivious algorithmic design; compact index construction; SNP-aware mapping by discounting dbSNP-specified mismatches; reporting of multiple mapping loci per read within an error threshold; parallel execution across multiple cores/processors; generation of ≈2 GB index for the human reference genome.