Graph analysis of fragmented long-read bacterial genome assemblies

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long-read genome assembly aims for perfect bacterial genome reconstruction.
• Fragmented assemblies persist, necessitating methods for improvement.
• Understanding limitations of current assemblers is crucial for genome refinement.

Purpose of the Study:

• To develop computational techniques for inspecting and improving fragmented bacterial genome assemblies.
• To assess if fragmentation is inherent or due to assembler limitations.
• To provide cost-effective strategies for bacterial genome finishing.

Main Methods:

• Analysis of assembly graphs to identify lost contig connections.
• Utilizing overlapping raw read paths to bridge fragmented contigs.
• Employing weighted Hamiltonian cycles for contig ordering.

Main Results:

• Recovered 45% of missing contig adjacencies in fragmented Canu assemblies.
• Hamiltonian cycle enumeration successfully suggested correct contig order in 50% of cases.
• Top-three contig order predictions were achieved in nearly all evaluated cases.

Conclusions:

• Computational analysis of assembly graphs can significantly refine fragmented bacterial genomes.
• The proposed methods offer a direction for finishing long-read assemblies with reduced experimental effort.
• Assembler limitations, rather than inherent impossibility, contribute to fragmentation in some bacterial genomes.