Benchmarking Long-Read Assemblers for Genomic Analyses of Bacterial Pathogens Using Oxford Nanopore Sequencing

Area of Science:

• Genomics
• Bioinformatics
• Microbial Pathogenesis

Background:

• Oxford Nanopore (ONT) sequencing offers long reads for complete bacterial genomes, but with higher error rates than Illumina short reads.
• Existing long-read assemblers require benchmarking for bacterial pathogen genomic analyses using ONT data.

Purpose of the Study:

• To benchmark the performance of six long-read assemblers (Canu, Flye, Miniasm/Racon, Raven, Redbean, Shasta) for bacterial pathogen genomics using ONT long reads.
• To evaluate assembler accuracy in reconstructing complete genomes and inferring key genomic features like antimicrobial resistance and virulence genes.

Main Methods:

• Benchmarking of six assemblers using simulated (mediocre and low-quality) and real ONT long reads from ten bacterial species.
• Evaluation of genome completeness, accuracy, antimicrobial resistance (AMR) profiling, virulence gene prediction, multilocus sequence typing (MLST), phylogenetic inference, and pan-genome analysis.

Main Results:

• Raven consistently produced the most robust and accurate complete bacterial genomes.
• Miniasm/Racon and Raven demonstrated strong performance in AMR profiling, MLST, phylogenetic inference, and pan-genome analyses, especially with real reads.
• Raven showed superior accuracy in virulence gene prediction, particularly with low-quality reads.

Conclusions:

• Raven is the most robust and accurate long-read assembler for bacterial pathogen genomics using ONT data, with Miniasm/Racon as a strong alternative.
• Assembler performance varies depending on read quality and specific genomic analysis task, highlighting the need for careful selection.