The parallelism motifs of genomic data analysis

Area of Science:

• Computational Biology
• Bioinformatics
• High-Performance Computing

Background:

• Genomic datasets are rapidly expanding due to decreased sequencing costs and accessible devices.
• Existing community databases facilitate data sharing, but large-scale computational resources are needed for complex analyses.

Purpose of the Study:

• To identify common computational patterns ('motifs') in high-performance genomics analysis.
• To inform parallelization strategies for memory- and compute-intensive genomic tasks.
• To highlight missing motifs in current parallelization frameworks.

Main Methods:

• Analysis of high-performance genomics problems including alignment, profiling, clustering, and assembly.
• Identification and comparison of computational patterns across different genomic analyses (single genomes and metagenomes).

Main Results:

• Several common computational motifs were identified for genomics analysis.
• Existing parallelization strategies may overlook essential patterns.
• Sorting and hashing were identified as critical, yet often missing, motifs for efficient parallelization.

Conclusions:

• Genomics analysis presents unique computational demands distinct from traditional scientific simulations.
• Incorporating motifs like sorting and hashing is essential for optimizing parallel architectures and software for high-performance genomics.
• Further development of programming support, libraries, and architectural designs is needed to meet these specific requirements.