I am working on a meta analysis of 18S amplicon data consisting of hundreds of samples. All the data has been demultiplexed and the paired end reads have been merged previously. I’m using deblur since it works with the merged reads data, however, for some samples I have reads in fasta format with no associated qual files. Would it work to just convert these fasta files into fastq using an arbitrary quality score of say, 30 or 40 for all bases in all reads? From reading the deblur paper, it seems like the quality scores aren’t needed for the deblur algorithm itself, but only for the initial quality filtering step. It also looks like the standalone version of deblur can accept fasta input, but the qiime 2 plugin doesn’t offer that option. I appreciate any input that can help me in creating a hack to get around this issue.
Good analysis of the situation. Your solution should work fine for deblur.
We have an issue to accept fasta as input in the future, but until then, any quality score should work. You might try setting them to 0 just so it's very obvious that the data isn't real.