In general, I'd agree with that. However, biological data can be huge and the least amount of processing you can get away with, the better.

I'd probably recommend validating the fasta file once, then setting it to "read-only". Perhaps additional checks to ensure it hasn't changed since validation might be in order. It can then be used multiple times with some reasonable degree of confidence about the data integrity.

Obviously, at this point, we don't know the source of the input, or even what it looks like, so validation requirements are purely guesswork.