Skip to content
Learn Motion
ExploreHow it worksMembership
Log in
Learn Motion

How AI Designs a Drug Molecule from Scratch

1Why Designing a Drug Molecule Is Hard2Turning Molecules into Something a Machine Can Read3How Generative Models Propose New Molecules4Scoring and Filtering the Candidates5Testing, Learning, and Improving the Design
How Generative Models Propose New Molecules

Proposing a molecule versus looking one up

1 / 3
The comparison on this page is the whole point of the chapter, so look at the two columns as two different kinds of limit. In library search, the limit is the collection: you can rank every compound you own, but the answer has to be in the drawer already. In a generative model, the limit is the encoding: the model can write any structure its representation can express, which is a much larger space, but it can only write in that language. That is why the fingerprint from the last chapter is out — you can score with it, but you cannot decode it back into atoms and bonds. And notice the word sample. The model is not picking from a shelf; it is making a series of probabilistic choices, which is why the same model asked twice gives you two different molecules.
0:00 / 0:00

Two ways to obtain a candidate

Library search

  • Input: a fixed, finite collection of existing compounds
  • Operation: score every entry, rank, keep the top ones
  • Boundary: the answer must already be in the collection
  • Output: a ranked subset of known molecules

Generative model

  • Input: a learned probability distribution over structures
  • Operation: sample a new structure from that distribution
  • Boundary: only the encoding's expressive range limits the output
  • Output: molecules that may never have been made or catalogued

Why sampling is the right word

The model holds probabilities, not a list. Producing a molecule means making a sequence of choices — which atom or fragment comes next, how it connects — where each choice is drawn according to what the model considers plausible. Because those draws are random, running the model twice on the same input gives two different molecules. This is a feature, not a defect: it is the only reason a single trained model can supply a whole pool of candidates instead of one guess.

The encoding constraint returns

A model can only generate in a language it can write. Fingerprints record which features are present but cannot be turned back into atoms and bonds, so they support scoring and similarity but not generation. Graph- and string-based encodings can be decoded, which is why they are the ones used when the task is to propose a structure rather than rank one.

Previous1 / 3Next

Learn Motion

Generate a course. Learn it properly.

Operated by Wuhan Daoyin Technology Co., Ltd.

Contact: [email protected]
Privacy PolicyTerms of Service

© 2026 Learn Motion