Coming back to post #59, because the follow-up matters more than the original answer.
Preclinical versus clinical: compounds with strong preclinical data can fail in clinical trials. Compounds with no human data are being used on theoretical grounds. The regulatory position reflects this asymmetry.