Drug discovery glossary – Learning Center

A

ADMET: ADMET stands for absorption, distribution, metabolism, excretion and toxicity — the properties that decide whether an active molecule can become a safe, effective drug.

Beyond rule of five (bRo5): Beyond rule of five (bRo5) describes orally active compounds that break Lipinski's guidelines — usually heavier than 500 Da — a region where many macrocycles sit.
Binding site: A binding site is the region of a protein where a ligand binds; its shape and chemistry determine which molecules can bind and how druggable the target is.

Chemical space: Chemical space is the set of all possible molecules; the drug-like part alone is estimated at around 10^60 compounds, far more than can ever be made.
Covalent drug: Covalent drugs form a chemical bond with their target protein, usually through a reactive "warhead" that attacks a nearby amino acid such as cysteine.

DiffDock: DiffDock is an open-source deep-learning docking method that uses a diffusion generative model to place a ligand on a protein without a predefined binding box.
Docking score: A docking score is a scoring function's estimate of how favourably a ligand pose interacts with a protein; it is useful for ranking but is not a measured binding affinity.
Drug discovery: Drug discovery is the process of finding a molecule that changes a disease-relevant biological target safely enough to be tested in people.

Fragment linking: Fragment linking joins two small fragments that bind neighbouring sites on a protein with a linker, aiming for a molecule that binds much more tightly than either fragment.

Hit, lead and candidate: A hit shows activity in a screen, a lead is a confirmed and improvable chemical series, and a clinical candidate is the single optimized compound taken into human trials.

Macrocycle: A macrocycle is a molecule containing a ring of twelve or more atoms; in drug discovery, macrocycles can bind large, flat protein surfaces that small molecules struggle to reach.
Membrane permeability: Membrane permeability is how readily a molecule crosses a cell membrane; assays such as PAMPA and Caco-2 measure it, and it is often the deciding property for macrocycles.
Molecular docking: Molecular docking predicts how a small molecule fits into a protein's binding site and estimates how well it binds, using a search algorithm and a scoring function.
Molecular fingerprint: A molecular fingerprint encodes a molecule's structural features as a fixed-length bit or count vector so that millions of molecules can be compared quickly.

Protein Data Bank (PDB): The Protein Data Bank is the global, free archive of experimentally determined 3D structures of proteins and nucleic acids, distributed in the PDB and mmCIF file formats.
Protein structure prediction: Protein structure prediction computes a protein's 3D shape from its amino-acid sequence; AI models such as AlphaFold and OpenFold now reach near-experimental accuracy for many proteins.

Scaffold hopping: Scaffold hopping replaces the core of an active molecule with a different chemical framework that keeps the key interactions, to gain new IP, better properties or easier synthesis.
SMILES: SMILES is a line notation that writes a molecule's structure as a short text string, such as CCO for ethanol, so software can store, search and exchange it.
Substructure search: Substructure search finds every molecule in a library that contains a given fragment or pattern, such as a ring system or functional group.

Tanimoto similarity: Tanimoto similarity compares two molecular fingerprints as shared features divided by total features, giving a score from 0 (nothing in common) to 1 (identical fingerprints).

Virtual screening: Virtual screening uses computers to rank large compound libraries so that only the most promising molecules are bought, made or tested.