Chemical features mining provides new descriptive structure-odor relationships

Abstract

An important goal in researching the biology of olfaction is to link the perception of smells to the chemistry of odorants. In other words, why do some odorants smell like fruits and others like flowers? While the so-called stimulus-percept issue was resolved in the field of color vision some time ago, the relationship between the chemistry and psycho-biology of odors remains unclear up to the present day. Although a series of investigations have demonstrated that this relationship exists, the descriptive and explicative aspects of the proposed models that are currently in use require greater sophistication. One reason for this is that the algorithms of current models do not consistently consider the possibility that multiple chemical rules can describe a single quality despite the fact that this is the case in reality, whereby two very different molecules can evoke a similar odor. Moreover, the available datasets are often large and heterogeneous, thus rendering the generation of multiple rules without any use of a computational approach overly complex. We considered these two issues in the present paper. First, we built a new database containing 1689 odorants characterized by physicochemical properties and olfactory qualities. Second, we developed a computational method based on a subgroup discovery algorithm that discriminated perceptual qualities of smells on the basis of physicochemical properties. Third, we ran a series of experiments on 74 distinct olfactory qualities and showed that the generation and validation of rules linking chemistry to odor perception was possible. Taken together, our findings provide significant new insights into the relationship between stimulus and percept in olfaction. In addition, by automatically extracting new knowledge linking chemistry of odorants and psychology of smells, our results provide a new computational framework of analysis enabling scientists in the field to test original hypotheses using descriptive or predictive modeling.

Fig 1. Building process of the database used for the study.

It was based on the Arctander’s Book and PubChem databases for determining a total of 74 olfactory qualities. Dragon software was used to obtain 82 physico-chemical properties of the 1689 molecules.

Fig 2. Word-cloud of the 74 studied olfactory qualities.

Fig 3. Rules and combinations of rules.

(a) Histogram showing the number of rules (x-axis; one to twelve) used to describe the 74 olfactory qualities (number of olfactory qualities, y-axis). (b) Example of the selection of the best rule or combination of rules based on the calculation of Euclidean distance from the “ideal” point (error “0”, recall “1” indicated as ● in the graph) for the olfactory quality jasmine. The number in parenthesis indicates the number of molecules described by this quality. The symbols (▲, ■, x, -) below the graph, indicates the different minSupp used for computation. The right panel indicates the combination of rules selected. (c) Chemical structure and name of the 29 molecules described as “jasmine”.

Fig 4. Quartiles and group distribution of the best rules describing the 74 qualities.

Quartile distribution was based on the Euclidean distance of a rule/combination of rules from the “ideal” point (error of “0”, recall of “1” indicated as ● in the graph). Quartiles are indicated by different colors (Q1 = blue, Q2 = pink, Q3 = grey, Q4 = orange). Group distribution (indicated as Group 1, 2, 3, and 4) was based on error rate (0.5) and recall (0.5). The 4 groups are separated by the dotted horizontal and vertical lines and indicated in the figure.

Fig 5. Validation studies in novel odorants from different available databases.

Bars correspond to means and error bars to the standard error to the mean (SEM). (a) Dravnieks dataset: odorants that fulfill the criteria for the rules woody (a.i), earthy (a.ii), and camphor (a.iii) from our model (Rule (1), black bars) were significantly more described by these respective qualities than odorants that did not fulfill these physicochemical rules (Rule (0), grey bars). (b) Licon et al. dataset: significant validation was observed for camphor. (c) Boelens & Haring dataset: validation of our rules was observed for woody (c.i) and balsamic (c.iv). (d) Keller et al. dataset: odorants that fulfill the criteria for rules associated with fruity were indeed perceived as more fruity for both concentrations of odorants (d.i: low concentration; d.ii: high concentration). Validation of the rules associated with the quality sulfuraceous was also observed (odorants that fulfill the criteria were perceived as more decayed at both low, d.iii, and high, d.iv, concentrations). ***p<0.001, **p<0.01, *p<0.05.