From molecule to medicine

A pharmacophore (as defined by IUPAC) is the set of shape and charge features a molecule needs to fit a biological target and trigger a response. It is an idea, a pattern of features, not an actual molecule. It explains how quite different molecules can fit the same receptor: they share the same pharmacophore.

Lipinski's rule of five (1997) predicts whether a compound is likely to work as a drug taken by mouth. It should have:

• No more than 5 hydrogen bond donors.
• No more than 10 hydrogen bond acceptors.
• A molecular weight below 500.
• A log P below 5 (log P measures how much the molecule prefers fat to water).

At most one rule may be broken. (All the numbers are 5 or multiples of 5, hence the name.) The rule does not apply to drugs given by injection or to natural products.

An ideal drug is non-toxic, stable, free of side effects, spreads evenly where needed, releases at a controlled rate, is affordable, and leaves the body easily.

Virtual screening filters billions of compounds in stages, each stricter than the last: ADME filters (absorption, distribution, metabolism, excretion) for drug-likeness; ligand-based methods (machine learning, pharmacophore searches); and structure-based methods, such as protein-ligand docking, which tests how each compound fits into the target protein.

Developing a drug is long and risky: thousands of compounds at the discovery stage narrow down to a handful in development and perhaps one approved drug, over many years. Key terms: high-throughput screening (testing huge numbers quickly), hits (compounds that show activity), false positives, a lead compound (active and worth developing further), a new chemical entity, and off-target activity (unwanted effects on other proteins).