In short
A molecular descriptor is a number computed from a structure: molecular weight, an estimated log P, a polar surface area, counts of rings, rotatable bonds and hydrogen-bond donors. Descriptors relate well to physical behavior — volatility, solubility, substantivity on fabric — because those depend on size and polarity. They relate much less reliably to odor quality, which depends on receptors and on details such as stereochemistry that most 2D descriptors cannot see.
What you will learn
- Write and read SMILES strings for common odorants
- Compute standard 2D descriptors with an open-source toolkit
- Test whether a descriptor correlates with your own tenacity observations, and interpret the result cautiously
- Explain why descriptors relate better to physical behavior than to odor quality
Before this lesson: Fragrance Calculations in Python, Checked, Reading a Molecular Structure, Predicting Odor from Structure
Before you start
You need to read a structure, the Molecular Perfumery track's predicting odor from structure, and the Python of the previous lesson. This uses RDKit, an open-source toolkit, outside the RUŌOD Lab app, which does no molecular calculation.
SMILES
SMILES writes a structure as text. Linalool is CC(C)=CCCC(C)(O)C=C; vanillin COc1cc(C=O)ccc1O. Lowercase letters are aromatic atoms, numbers close rings, parentheses are branches. Copy SMILES from PubChem rather than writing them by hand until you can check your own.
Computing descriptors
from rdkit import Chem
from rdkit.Chem import Descriptors
for name, smi in [("linalool", "CC(C)=CCCC(C)(O)C=C"),
("vanillin", "COc1cc(C=O)ccc1O")]:
m = Chem.MolFromSmiles(smi)
print(name,
round(Descriptors.MolWt(m), 2), # molecular weight
round(Descriptors.MolLogP(m), 2), # Wildman–Crippen log P estimate
round(Descriptors.TPSA(m), 2), # topological polar surface area
Descriptors.NumHDonors(m))Compare your molecular weights with PubChem's; they should agree to two decimals. Estimated log P values differ between methods by a fraction of a unit or more, so name the method whenever you report one.
What descriptors can tell you
- Volatility, roughly. Within a chemical family, molecular weight and hydrogen bonding track vapor pressure, which governs evaporation. Across families the relation is loose: a hydrogen-bonding alcohol is far less volatile than a hydrocarbon of the same mass.
- Partitioning. log P relates to how a material divides between water and oils, which matters for functional products and deposition on skin and fabric.
- Screening. In odorant design, descriptor filters narrow candidates before synthesis.
What they cannot
- Odor quality. Structurally similar molecules can smell very different and dissimilar ones alike, as the musk and ambergris case studies show.
- Stereochemistry. Enantiomers share 2D descriptors; carvone's two forms are the classic counterexample.
- Mixtures. A descriptor describes one molecule; perfumes are mixtures.
Open questions
How much of a material's blotter tenacity is explained by descriptors alone, and how much needs measured vapor pressure and interactions with paper? A dataset of descriptors beside carefully timed blotter lifetimes, for a few dozen materials, would be small enough for one person and genuinely informative.
Practical exercise: Descriptors against your own blotter data
- Choose ten single-molecule materials you own, spanning top to base. Copy each SMILES from PubChem and record the record's address.
- Compute molecular weight, log P (name the method), TPSA and hydrogen-bond donors.
- Measure each material's blotter lifetime at 10% in ethanol: dip, then smell blind (coded strips) at fixed intervals until gone. Record room temperature.
- Plot lifetime against molecular weight and against log P. Calculate a rank correlation for each.
- Write what the correlations show for these ten materials, and why they do not establish a rule.
You have done this well if every SMILES has a source, your molecular weights match PubChem's, lifetimes were measured blind under recorded conditions, you used a rank correlation for a small sample and said why, and your interpretation names at least one material that breaks the pattern and explains it.
Common mistakes
- Hand-written SMILES with errors. Check every one against a database.
- Comparing log P values from different methods.
- Reading descriptor correlations as odor rules.
Safety
Blotter work only, in a ventilated room. Read each SDS and follow the safe handling lesson.
Key terms
- SMILES
- A line notation that writes a molecule's structure as text, such as CC(C)=CCCC(C)(O)C=C for linalool. Stereochemistry can be included with @ and / \ signs.
- Molecular descriptor
- A number calculated from a structure, such as molecular weight, an estimated log P or the count of hydrogen-bond donors, used to compare molecules or as input to a model.
- Topological polar surface area
- An estimate of the surface area contributed by polar atoms, computed from fragment contributions. Higher values mean more polar molecules.
Knowledge check
Written and reviewed by the RUŌOD Lab team. This is general education, not legal, regulatory, medical or safety advice. Example dilutions and exercises are for learning, are not validated products, and have not been safety-assessed. Completing a lesson confers no qualification. How we check these lessons.