gmiclotte
commented
3 years ago Hello
Firstly, Brownie and Jabba were not developed with metagenomics in mind. For example, the Brownie+Jabba approach does not perform well when there is a (local) low coverage of short reads and this can result in a significant loss of long read coverage (see A comparative evaluation of hybrid error correction methods for error-prone long reads ). Instead, you could have a look at metagenomics-aware assemblers, such as OPERA-MS( repository, paper ).
Secondly, to answer your question: a single de Bruijn graph should be built from all the reads. Brownie and Jabba are intended for DNA sequencing data, and the graph construction algorithm correctly handles both the forward and reverse complement reads.
Kind regards Giles Miclotte
I'm wondering what would be the most appropriate method to go about using Brownie on paired-reads from metagenomic samples? I would like to use both the forward and the reverse reads for long-read error correction with Jabba, but I can't imagine how to do this without using two sets - first building a graph with the forward reads with Brownie, then building a graph with the reverse reads with Brownie.
I'm wondering if there's a way to circumvent this. If anyone has a clearer/more informed approach it'd be appreciated!