Objectives are incomplete, data are biased, physical evaluation is expensive and biological validity cannot be reduced to a simple verifier.
A possibility worth exploring ●
AI-Guided Molecular Search
Use generative models, active learning and physical or experimental checks to navigate huge molecular design spaces.
The possibility
Use generative models, active learning and physical or experimental checks to navigate huge molecular design spaces.
The opportunity already exists: the key is closing the loop between proposal, prediction, uncertainty and experiment.
Better search and optimization methods could improve candidate efficiency, but experimental evidence remains essential.
What to keep in mind
What stands between here and there?
Do not describe P versus NP as the blocker for drug discovery; this is an indirect opportunity relation.
P versus NP
Combinatorial complexity helps explain search difficulty, but resolving P versus NP is not a prerequisite or a sufficient solution for molecular discovery.
Understand the question →Ideas it sparks
Ways to start thinking today.
These are prompts for curiosity—not roadmaps or promises.
Molecular Search Evidence Loop
Propose candidates, estimate uncertainty, select the most informative experiments and feed measured results back into the model.
Why it is thinkable now
Explore the idea →Yes in narrowly defined chemical or materials domains.
Sources and uncertainty 2
Reducibility among Combinatorial ProblemsRichard M. KarpSource verifiedOptimal Molecular Design: Generative Active Learning Combining REINVENT with Precise Binding Free Energy Ranking SimulationsHannes H. Loeffler, Shunzhou Wan, Marco Klähn, Agastya P. Bhati, Peter V. CoveneySource verified
AI summarized and source-backed · Qualified expert review pending.