why do alignment before classification?

Hi guys!

I just have a doubt that why we conduct alignment before classification, as before I consulted a similar question to a RDP staff and he told me that “The alignment is NOT needed for RDP Classifier to make assignment. It would be better off just use your raw reads for the classification.” Because after alignment and filtering in mothur, the sequence length will reduce and then the result of classification will not be so accurate as using the raw reads. So should we firstly reduce sequence errors, the do classification and at last do alignment? Thank you very much!

Technically speaking they are correct - you do not need an alignment to perform the classification. However, if you don’t align your sequences, you don’t know that they overlap the same region and you can’t trim them to the same alignment coordinates, which is important since the 16S gene does not evolve uniformly over its length. Sure the trimmed sequences will be shorter and not as accurate to classify, but then again, it’s hardly “accurate” to compare the classifications of 200 bp sequences to 400 bp sequences.