An “artificial nose” reads odors as electrical signals. To develop its materials, researchers need to know which molecules capture odor compounds and what electrical changes occur after capture. If calculations could narrow down promising matches before experiments, the search for materials could become easier.
MORE-Q, published by Chen and colleagues in 2025, is a quantum-chemical dataset designed for that purpose. It combines 102 molecules associated with body odor with 18 artificial receptor molecules, documenting their structures and electronic behavior. Rather than comparing the performance of finished sensors, the study creates a foundation for designing sensors and understanding how they work.
Not the receptors found in the nose
The receptor molecules here are not the olfactory receptor proteins found in cells in the human nose. They are artificial molecules inspired by mucin, a component of mucus, with various chemical groups attached to a sugar-based scaffold. Eighteen is neither the number of people nor the number of receptor types in the nose.
Changing the groups around the sugar allows properties such as molecular rigidity, hydrophobicity, and polarizability to be adjusted. Some of the groups in this study help the molecules adhere to graphene. Graphene is a highly electrically conductive material consisting of carbon atoms arranged in a thin sheet.
In the sensor envisaged by the researchers, receptor molecules interact with odor compounds, and those changes are coupled to graphene’s electrical properties. It is therefore important to study not just the odor compound and receptor molecule, but also the supporting surface.
Combining 102 molecules with 18 partners
The odor compounds were not selected by collecting fresh human sweat. The researchers cross-referenced records of volatile substances reported from the human body with an existing dataset containing odor-perception ratings, selecting 102 skin-related molecules. The molecules studied contain carbon, hydrogen, oxygen, nitrogen, and sulfur.
Multiplying 102 by 18 gives 1,836 combinations. However, even the same two molecules can differ in stability depending on their relative orientation and attachment site. The researchers explored various arrangements of each odor molecule around each receptor molecule.
After removing overly similar arrangements, 23,838 configurations remained. This does not mean that more than 20,000 different odors were studied. It is the number of different attachment configurations examined across the 1,836 pairs.
The data are divided into three stages. The first covers the properties of individual odor and receptor molecules. The second covers their properties when paired. The third selects one energetically favorable configuration from each pair and examines all 1,836 pairs on graphene. The progression is from isolated molecules toward combinations closer to device materials.
Exploring many configurations, then calculating selected ones in detail
Quantum-chemical calculations derive molecular energies, charge distributions, and other properties from the relationships between atomic nuclei and electrons. But the more atoms and configurations involved, the more time the calculations take.
The researchers therefore first used a relatively inexpensive computational method to optimize structures and explore many configurations. This lighter method was also used to calculate the properties of the 23,838 configurations. They then used density functional theory, a more detailed electronic-structure approach, for selected systems including the 1,836 pairs. Not every configuration was calculated at the same level of accuracy.
The records include molecular shapes, energies, atomic charges, polarizability, and other properties, with up to 39 properties depending on the stage. For systems on graphene, changes before and after odor-molecule adsorption were also calculated. The data are publicly available for other researchers to retrieve and analyze.
The researchers also examined whether the 1,836 selected configurations omitted substantial portions of the properties represented in the larger set. When several properties were reduced to two axes for comparison, the selected configurations were distributed broadly across the region occupied by the original data. The authors consider the reduced set to represent some of the original diversity. This is not a guarantee that every possible configuration has been covered.
Does stronger attachment mean a larger electrical signal?
One property highlighted in the study is the “work function.” In simple terms, it indicates the energy needed to remove an electron from a material. Because it changes when an odor molecule attaches, it offers clues to a sensor’s electrical response.
In the calculations, odor-molecule adsorption increased the work function for some combinations and decreased it for others. The distribution of atomic charges within receptor molecules also changed, showing how interactions with odor molecules alter local electronic states.
However, there was no simple correlation between adsorption energy, which describes how strongly a molecule attaches, and the magnitude of the work-function change. Nor was there a simple correspondence between changes in the overall charge of graphene and changes in its work function.
Within the scope of these calculations, then, selecting strongly binding pairs does not necessarily produce larger electrical changes. The authors view the lack of a single relationship constraining these properties as offering freedom to tune several properties in sensor design. Rather than ranking materials by binding strength alone, there is value in exploring candidates from several perspectives.
Connecting materials discovery with human odor perception
Calculated work functions do not, however, directly establish a real sensor’s detection limit or its accuracy in identifying mixed odors. Nor did this paper test the accuracy of human body-odor diagnosis using a new device.
The elements and receptor types covered are limited, and detailed calculations on graphene focus mainly on one selected configuration for each pair. The authors themselves note that exploring receptor orientations and conformations more broadly could change the results. More precise computational methods and larger numbers of configurations could improve the work, but would also increase computational costs.
One proposed future use is to train machine-learning models on the public data to rapidly predict properties of combinations that have not yet been calculated in detail. The authors also envisage combining the dataset with generative models to propose new receptor-molecule structures.
They further propose connecting data on molecular binding and electronic changes with data on how people rate odors. This is not a conclusion that human perception can be read directly from changes in artificial materials. It is a future direction for advancing understanding of olfaction by linking different kinds of information.
Summary
MORE-Q is a dataset that brings together odor molecules, artificial receptor molecules, and electrically conductive graphene, enabling comparisons of structural and electronic changes when molecules attach.
Its key contribution is to make strong binding and large electrical changes separately investigable. The authors hope these foundational data will support the selection of materials with an understanding of their mechanisms, providing a basis for designing new artificial noses.
References
Chen, L., Medrano Sandonas, L., Traber, P., Dianat, A., Tverdokhleb, N., Hurevich, M., Yitzchaik, S., Gutierrez, R., Croy, A., & Cuniberti, G. (2025). MORE-Q, a dataset for molecular olfactorial receptor engineering by quantum mechanics. Scientific Data, 12, 324. DOI: 10.1038/s41597-025-04616-6 / Full original article / Dataset described in the paper. Original article: © The Author(s) 2025, CC BY 4.0.
