Long-read nanopore metagenomics for reconstruction of bacterial genomes
Sam: We know that faecal microbial transplants work – but why? What changes happen in the community that allow these transplants to be successful? This is what Sam’s team are looking at.
Their pipeline reticulatus is a GPU-accelerated long-read metagenomics pipeline, involving Guppy basecalling, MetaFlye assembly, polishing with racon, medaka, and pilon. GitHub: samstudio8/reticulatus.
With file parsing, GPU acceleration, and ‘a whole lot of other tricks’, they massively reduced racon iteration time to ~10min on the PromethION data (from the original 600 min using the CPU).
‘Not every bioinformatician gets to go to the algorithm ball, somebody has to stay at home and write the pipelines’.
Sam asked us to ‘spot the difference’ between the pre-and post-FMT microbial profiles: one genus completely disappears, and others are present only post-transplant.
Long-read assembly and polishing can yield Q50 genomes.