Taming large-scale genomic analyses via sparsified genomics

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genomic sequence comparison is crucial for biomedical research.
• Current methods struggle with the vast and growing volume of sequencing data.
• Need for efficient and scalable computational tools in genomics.

Purpose of the Study:

• Introduce "sparsified genomics" to accelerate genomic data processing.
• Enable faster and memory-efficient analysis of genomic sequences.
• Maintain accuracy comparable to traditional methods.

Main Methods:

• Systematically excluding a large number of bases from genomic sequences.
• Applying sparsification to accelerate read mapping (e.g., minimap2).
• Utilizing sparsified sequences for containment search and taxonomic profiling (e.g., CMash, KMC3, Metalign).

Main Results:

• Accelerated minimap2 performance by 2.57-6.28x across different read types.
• Achieved 72.7-75.88x faster and 723.3x more storage-efficient containment searches.
• Enabled 54.15-61.88x faster and 720x more storage-efficient taxonomic profiling of metagenomic samples.
• Maintained comparable accuracy and reduced index size by 2x.

Conclusions:

• Sparsified genomics offers substantial speed and storage advantages for various genomic analyses.
• This approach is broadly applicable, enhancing tools for read mapping, sequence searching, and microbiome discovery.
• Provides a scalable solution to handle the exponential growth of genomic data.