deGSM: Memory Scalable Construction Of Large Scale de Bruijn Graph

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• De Bruijn graphs are crucial for genome sequence representation and analysis.
• High-throughput sequencing (HTS) generates massive datasets, demanding efficient large-scale de Bruijn graph construction.
• Existing methods struggle with the memory requirements of terabase-pair scale genome data.

Purpose of the Study:

• To introduce de Bruijn Graph Constructor in Scalable Memory (deGSM), a lightweight, parallel approach for constructing de Bruijn graphs.
• To enable the efficient handling of extremely large genome sequences (terabase-pair level).
• To provide a scalable and memory-efficient solution for modern genomics studies.

Main Methods:

• deGSM constructs the Burrows-Wheeler Transformation (BWT) of de Bruijn graph unipaths in constant RAM.
• It then transforms the BWT back into the original unitigs.
• The approach is designed for parallel processing and scalability.

Main Results:

• deGSM successfully processed very large genome datasets, including 305 Gbp contigs, 1.1 Tbp scaffolds, and a 9.7 Tbp HTS dataset on a standard machine.
• Construction speed was comparable or faster than state-of-the-art methods.
• Demonstrated high scalability and efficiency for large-scale genomics.

Conclusions:

• deGSM offers a memory-efficient and scalable solution for constructing large de Bruijn graphs.
• Its performance makes it suitable for handling massive genome datasets generated by HTS.
• The tool has significant potential for advancing large-scale genomics research and analysis.