SWAP-Assembler

SWAP-Assembler performs scalable parallel de novo genome assembly of large short-read sequencing datasets using the Small World Asynchronous Parallel (SWAP) model.

Key Features:

• Scalability: Scales to thousands of computational cores with reported use of up to 2048 cores for large datasets.
• Parallelization: Executes fully parallelized operations to resolve computational interdependencies during assembly, enabling concurrent processing across cores.
• SWAP and MSG frameworks: Implements the Small World Asynchronous Parallel (SWAP) model and a multi-step bi-directed graph (MSG) framework to manage distributed computation.
• Graph-based assembly: Treats genome assembly tasks as edge-merging operations on de Bruijn graphs within a semi-group formulation.
• Graph cleaning and contig extension: Incorporates graph cleaning and contig extension processes to produce high-quality contigs with superior N50 sizes and low error rates, demonstrated on datasets such as Fish and Yanhuang.
• Performance: Reported to outperform parallel assemblers including ABySS, Ray, and PASHA in scalability and contig quality and assembled the Yanhuang dataset in 26 minutes using 2048 cores.

Scientific Applications:

• Large-scale genomic projects: Enables assembly of massive sequencing datasets that exceed the capabilities of traditional assemblers.
• De novo genome assembly: Produces long contigs and high N50 values for de novo reconstruction of genomes from short reads.
• Metagenomic studies: Applies to metagenomic assembly of complex mixed-community sequencing data at large scale.
• Complex organismal genomics: Supports assembly of genomes with complex structures that require extensive computational resources.

Methodology:

Implements the Small World Asynchronous Parallel (SWAP) model and a multi-step bi-directed graph (MSG) framework to perform fully parallelized edge-merging operations on de Bruijn graphs within a semi-group formulation, combined with graph cleaning and contig extension.