I would like to ask if it is possible to create fastq files from filtered (DADA2) feature tables +rep seqs. In SRA metagenomic submission it is stated that there should be no human host sequences in the submitted sequences, but all unfiltered metagenomic sequences, even 16S, usually have a percentage of host chimeras. I could filter chimeras separately on my raw data, but then the files i will submit will differ from those i used in the analyses and there will a problem with the reproducibility of the results i guess.
In theory, it is possible, but the quality scores can't be restored at this point, and you will need to write mock values instead. I did it once for fun, but I would not submit such files to an online repository.
I wouldn't worry about chimeric sequences with host DNA in 16S rRNA gene amplicons because they're very rare. Moreover, they are short and can't be assembled. Just upload your raw 16S rRNA gene amplicon reads. But upload cleaned from host DNA human samples if you are working with shotgun metagenomic samples.