It is fine to use dada2 for initial denoising — but if performing a picrust analysis you must then perform closed-reference OTU picking agains the picrust database using vsearch, as described in the thread that you linked to. Since you mention dada2, I just want to make sure that you saw the vsearch step in that linked protocol.
Check out this thread, it looks like you are having the same error:
The error seems fairly clear: there is a formatting issue with your metadata file (i.e., 97_otu_taxonomy.txt) that needs to be fixed.
Note that biom-format and picrust are external software and we cannot really provide support for them on this forum, though we will do what we can to help.
Let me know if fixing the header line in your metadata file helps!