Comparison of short and long read metagenomic assembly on mice faecal samples

Shotgun metagenomics enables the reconstruction of complex microbial communities at a high level of detail. Such an approach can be conducted using both short-read and long-read sequencing data, as well as a combination of both. To assess the pros and cons of these different approaches, we used 22 faecal DNA extracts collected weekly for 11 weeks from two respective lab mice to study seven performance metrics over four combinations of sequencing depth and technology: i) 20 Gbp of Illumina short-read data, ii) 40 Gbp of short-read data, iii) 20 Gbp of PacBio HiFi long-read data, and iv) 40 Gbp of hybrid (20 Gbp of short-read + 20 Gbp of long-read) data. No strategy was best for all metrics, but instead, each one excelled across different metrics. The long-read approach yielded the best assembly statistics, with the highest N50 and lowest number of contigs. The 40 Gbp short-read approach yielded the highest number of refined bins. Finally, the hybrid approach yielded the longest assemblies, and the highest mapping rate to the bacterial genomes. Our results suggest that while long-read sequencing significantly improves the quality of reconstructed bacterial genomes, it is more expensive and requires deeper sequencing than short-read approaches to recover a comparable amount of reconstructed genomes. The most optimal strategy is study-specific, and depends on how researchers assess the tradeoff between the quantity and quality of recovered genomes.