Most downstream analyses require you to have multiple samples (really any dataset with less than 2 samples is going to be very limited) - so I encourage you to run q2-vsearch's dereplicate-sequences action! As well, you'll need at least a feature table for most downstream steps, so that's another reason to not skip this step.