News Release

预期化合物的气味

Peer-Reviewed Publication

American Association for the Advancement of Science (AAAS)

Anticipating the Aroma of a Chemical Compound

video: In a new paper published in Science, IBM Research presents the results of the crowdsourced DREAM competition that show it is possible to accurately predict and reverse-engineer perceptual attribute values for individual molecules. This material relates to a paper that appeared in the Feb. 24, 2017, issue of Science, published by AAAS. The paper, by A. Keller at The Rockefeller University in New York, N.Y., and colleagues was titled, 'Predicting human olfactory perception from chemical features of odor molecules.' view more 

Credit: Rockefeller University and IBM Research

通过众包竞争所收集的数据证明,通过给定的结构,准确预测某分子的气味是可能的。由梦想嗅觉预测挑战(DREAM Olfaction Prediction Challenge)所产生的这些结果不仅可被用于预测某特定分子的气味特性,而且还能创建新的代表人类能感受的品种繁多的气味与香气的分子。预测气味可能颇为困难,部分原因是某些具有非常相似化学结构的分子可被人类区分开来,然而某些结构迥异的分子却能产生几近等同的气味。为了更好地了解形成气味的分子特征,Andreas Keller和Leslie Vosshall收集了来自49个人的数据,这些人对476种结构和感觉上不同的分子进行了扼要描述。这些数据被呈交给了22个小组,这些小组争相创建最准确的气味预测模型、研发预测每种气味品质与某个给定分子间关系的算法。7种总体上能被最佳预测的气味属性有:气味强度、蒜味、令人愉悦的气味、甜味、果味、香料味和“糊味”。在没有建模时,某些香气需要数百种化合物才能得到最佳预测;然而,某些由DREAM项目所生成的模型只根据若干特性就能够准确地预测气味。例如,强度是本预测研究中最复杂的特性,它需要有15种特征;相比之下,而且有点意料之中的是,鱼的独特而且强烈的气味可根据一种特征的出现而被预测出来。总的来说,这些模型可成功地预测本研究中所分析的19种语义描述符中的8种。

###


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.