The Structure–Odour Relationship: Can We Predict What a Molecule Will Smell Like?

Can Molecular Structure Predict Smell? Understanding Structure–Odour Relationships in Perfumery

Explore the science of structure–odour relationships (SOR) and whether we can predict how a molecule will smell based on its chemical shape, size, and function.

Introduction: A Holy Grail in Fragrance Science

Perfumery is one of the last creative sciences where trial and error still dominates. Despite centuries of exploration, predicting what a new molecule will smell like — or whether it will smell at all — remains one of the most elusive challenges in olfactory science.

The field of structure–odour relationships (SOR) seeks to bridge that gap. It aims to answer a deceptively simple question: Can we predict odour from molecular structure? In this article, we explore the core theories, challenges, and real-world applications of SOR in modern fragrance chemistry.

1. What Is Structure–Odour Relationship (SOR)?

SOR refers to the study of how a molecule’s physicochemical structure relates to its perceived odour. This includes:

Functional groups (e.g., alcohols, ketones, esters) Molecular weight and volatility 3D shape and conformational flexibility Lipophilicity (LogP) and water solubility Electronic distribution and charge density

The goal is to correlate these properties with specific scent descriptors (e.g. “fruity,” “musky,” “green”) or threshold intensity.

2. Early Theories: Shape and Vibration

Two major early hypotheses dominated SOR:

▪ Shape Theory (Lock and Key Model)

Proposed that odour molecules fit into olfactory receptors like a key fits a lock — the better the fit, the stronger the scent. This aligns with how enzymes recognise substrates.

Issue: Molecules with very similar shapes can smell completely different — or not at all.

▪ Vibration Theory (Turin’s Hypothesis)

Suggests that odour detection is linked to molecular vibrations at specific energy levels (infrared frequencies), rather than shape. The receptor acts as a spectrometer, detecting vibrational spectra.

Issue: Experimental support is mixed; many synthetic molecules contradict the model.

Modern consensus holds that no single mechanism explains odour. Instead, SOR likely depends on a complex interplay of binding geometry, receptor dynamics, and downstream neuronal activation.

3. The Challenge of Odour Prediction

SOR is notoriously difficult because:

Odour is subjective — influenced by genetics, culture, language, and context Some molecules have multiple odours depending on concentration Small structural changes can yield radically different perceptual results Many odorants bind to multiple receptors, and many receptors respond to multiple molecules Olfactory receptors (ORs) belong to a huge gene family (~400 in humans), many with unknown ligands

These factors make it hard to create simple, one-to-one predictive rules.

4. Computational Modelling and AI in SOR

Despite these challenges, machine learning is now being used to build predictive models of odour perception.

▪ QSOR (Quantitative Structure–Odour Relationships):

Computational models use molecular descriptors (e.g., shape, polarity, electronic features) to classify molecules into scent categories.

▪ Graph neural networks (GNNs)

AI models trained on large olfaction datasets (e.g. the DREAM Olfaction Prediction Challenge) can now predict odour labels with moderate accuracy.

▪ OdorMap and SMILES-based models

Molecular strings (SMILES) are converted into feature vectors to train scent prediction algorithms, clustering compounds by perceptual similarity.

These models are still developing, but show promise for:

Virtual screening of novel aroma chemicals Ingredient substitution Fragrance allergen reduction Enhancing synthetic biology libraries for aroma compound production

5. Structure and Threshold: Why Some Molecules Are Smell-Free

Not all molecules are odorous — even if they’re volatile.

Odour threshold depends on:

Molecular binding affinity to receptors Accessibility of the receptor (nasal mucus barrier) Molecular size and diffusion coefficient Metabolic degradation before binding

Some compounds may have a similar shape to a known odorant, but fail to bind effectively — or bind to silent receptors. Conversely, trace-level contaminants in a fragrance may have extremely low thresholds, making them disproportionately impactful.

6. Practical Use of SOR in Fragrance Development

While we can’t fully predict odour yet, SOR guides real-world formulation in several ways:

Musks: Macrocyclic musks with 14–18 atoms tend to have sweet, powdery tones. Structure alterations yield variation in persistence and tenacity. Green notes: Unsaturated aldehydes with 6–10 carbon chains (e.g., (Z)-3-hexenal) deliver grassy, fresh accords. Citrus notes: Compounds with a cyclohexene or terpene backbone (e.g., limonene, citral) are frequently fruity or lemony. Sulphur-containing molecules: Small changes can shift scent from savoury to putrid — e.g., thiols, mercaptans, isothiocyanates

Skilled perfumers use both structure and precedent to design new molecules or blend substitutes — while still relying on olfactory evaluation.

Conclusion: Structure Is Only Part of the Story

While structure–odour relationships offer powerful insights, the complexity of smell perception resists easy answers. Molecules are more than their formulas — they are shaped by interaction, biology, concentration, and memory. The future of predictive perfumery lies in data-driven models, receptor biology, and AI-assisted design, but the nose remains the final judge.

SKD Pharmaceuticals embraces both tradition and innovation in fragrance science. By understanding SOR principles, we select ingredients not just for regulatory fit or supply chain availability — but for their true olfactory character. Our private label fragrance offerings are shaped by chemistry, verified through sensory precision, and crafted for performance in the real world.