There are some non-ML methods, like for example: https://en.m.wikipedia.org/wiki/Lipinski%27s_rule_of_five

IANAdrugologist but that rule is about bioactivity rather then psychoactivity. Open to correction though.

you're correct. it doesn't help a lot with predicting psychedelic activity other than to act as a negative filter for unlikely compounds.

Yes, it's just one piece of the puzzle. The op might also be interested in David E Nichols' research into structure-activity relationships of psychedelics. See: https://web.archive.org/web/20080704190159/http://www.heffte...

My chemical intuition says that yes, it could be done. However, I imagine psychedelic states are much more neurologically complex and molecules more chemically complex than the analagous experiences vis-à-via scent (although we certainly do tie smells to memories and other secondary and tertiary experiences).

If we can also predict side effects, maybe we could find the perfect drug ?

How would you quantify that?

Great question, one could use existing known chemicals as a starting point. There could be a potential to use fMRI readings on a model organism in realtime to generate data.

Compelling. I wonder what else this could be applied to in addition to psychedelics? Anti-anxiety and other sensory affecting drugs?

If you wanna get Black Mirror-esque, perhaps a Soma-like medication from Brave New World (essentially pacifies/zombifies you by creating endless bliss) could be made. Or the "bliss" drug episode of Doctor Who.

Sentiment analysis of trip reports of known compounds on erowid?

Shulgin gives ratings to compounds in PiHKAL. Those could be used as well.