Researchers study how data moves in DNA mapping tools
A study identifies how different software tools use memory and hardware during genomic analysis.
Read mapping algorithms are a critical bottleneck in genomic analysis pipelines. This problem grows as sequencing technologies advance and datasets expand.
Researchers analyzed several tools including MashMap, wfmash, Minimap2, NGMLR, LRA, Bowtie2, BWA-MEM, StrobeAlign, and minibwa. They looked at whole-genome alignment, long reads, and short reads.
The study used the Intel VTune Profiler for memory analysis. They also used the DAMOV-SIM simulator to see data movement in two different cache structures.
Read mapping algorithms are tools used to align DNA reads. Memory-bound functions are processes limited by the speed of data movement. Cache structures are parts of a processor that store data.
Minimap2 is the fastest for long read scenarios. However, it falls behind in whole-genome alignment due to poor multithreading.
Some programs like LRA and BWA-MEM are highly memory-bound. Some tools do not fully utilize the multi-level cache hierarchy in modern multi-core CPU architectures.
Key facts
- Minimap2 is the fastest tool for long read scenarios.
- LRA and BWA-MEM are highly memory-bound.
- Some tools may benefit from running on PIM architectures.