Well I had this whole lovely post written and had to re-sign in and it was deleted… Anyway, in short:
I would like to see additional quality checking in/before make.contigs. If you have fully overlapping reads and are using make.contigs to error correct, wouldn’t it make sense to filter out low quality reads before making contigs with them? Not the insert option, but more similar to the qaverage/qwindow options in trim.seq, but with the ability to handle paired reads and chuck both forward and reverse reads if one doesn’t pass, then you’re not error correcting with overall low quality sequences.
I would also like to see a truncation option in screen.seqs. The premise: it’s been suggested as few as 75bp could work in identifying an organism with the 16s gene (http://www.pnas.org/content/108/Supplement_1/4516). It would need to be tested further (and I personally wouldn’t go as short as 75bp), but why use/process 200bp of sequence if 175bp could give you the identification? 25bp X Nmillion seqs = a lot of processing that could be saved.