The peanut is an interesting example. I think projects are underway to produce modified varieties that lack the allergens which people tend to react to.

The assumption that we can better determine toxicity with our current understanding of human biology than thousands of years of natural selection seems questionable, but peanuts are certainly a good lower bound on selection's ability.

I also don't have much confidence that the parties responsible for safety testing are particularly reliable, but that's a loose belief.