The fundamental units of olfactory perception are discrete 3D structures of volatile chemicals that each interact with specific subsets of a very large family of hundreds of odorant receptor proteins, in turn activating complex neural circuitry and posing a challenge to understand. We have applied computational approaches to analyze olfactory perceptual space from the perspective of odorant chemical features. We identify physicochemical features associated with ~150 different perceptual descriptors and develop machine-learning models. Validation of predictions shows a high success rate for test set chemicals within a study, as well as across studies more than 30 years apart in time. Due to the high success rates, we are able to map ~150 percepts onto a chemical space of nearly 0.5 million compounds, predicting numerous percept–structure combinations. The chemical structure-to-percept prediction provides a system-level view of human olfaction and opens the door for comprehensive computational discovery of fragrances and flavors.

中文翻译：

---

对人类嗅觉化学空间的全系统理解

嗅觉感知的基本单位是挥发性化学物质的离散 3D 结构，每种结构都与数百种气味受体蛋白的非常大家族的特定子集相互作用，进而激活复杂的神经回路并构成理解挑战。我们已经应用计算方法从气味化学特征的角度分析嗅觉感知空间。我们识别与约 150 种不同感知描述符相关的物理化学特征，并开发机器学习模型。预测的验证表明，在一项研究中以及在相隔 30 多年的研究中，测试集化学品的成功率很高。由于成功率高，我们能够将约 150 个感知映射到近 50 万种化合物的化学空间，预测许多感知结构组合。化学结构到感知的预测提供了人类嗅觉的系统级视图，并为香水和香精的综合计算发现打开了大门。