Odor Identification Quantitatively by Odorant Receptor Vectors

doi:10.21203/rs.3.rs-1599106/v1

Download PDF

Research Article

Odor Identification Quantitatively by Odorant Receptor Vectors

https://doi.org/10.21203/rs.3.rs-1599106/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Background: Odors are perceived by individuals through sensation, memory, experience, etc., and mainly subjective and qualitative evaluation has been performed until now. On the other hand, since Buck et al. succeeded in cloning the odorant receptor in 1991, researchers have shown interest in olfaction system, and molecular biology researches on olfaction system have begun actively. However, no objective evaluation has been made on the identification or similarity of odors. As for the odorant sensors, an odor is identified by using quartz oscillators or semiconductors and measuring a change in characteristics caused by odor molecules adhering to the sensor. In this study, we attempted to quantitatively express the direct association between odorant receptors and odor molecules at each concentration. The degree of similarity evaluated by taking inner product of the odorant receptor vectors was verified by comparing with the results by sensory evaluation tests published in past literatures in the form of correlation coefficient. A multi-layered artificial neural network whose input was the elements of prescribed odorant receptor vector was also introduced to identify odorant molecules in the output layer.

Results: The inner product of proposed odorant receptor vector might evaluate the degree of similarity quantitatively among various odorants. Further, it might be possible to identify odorant molecules by the artificial neural network using EC50 of receptors to various odorants.

Conclusions: In this paper, we proposed a vector consisting of the EC50 of odorant receptors and compared the inner product of the vector with the results of sensory evaluation tests based on the responses of odorant receptors to odorant substances at various concentrations. It was shown that not only the correlation between each odorant substance but also the correlation between the inner product of the odorant substance receptor vector and the results of the sensory evaluation test can be quantitatively evaluated. It was also shown that the EC50 of each odorant receptor can be placed in the input layer of an artificial neural network, and that an artificial neural network with odor molecules as the output layer can discriminate odors.

Odor

Odorant receptor

Correlation estimate

Similarity analysis

Identification

Artificial neural network

Human sensing systems are known to divide into five senses; vision, audition, tactility, olfaction, and gustation. Visual, auditory, and tactile sensations have a high degree of vitality and can be handled physically, whereas olfaction and gustation often relate to ensuring the quality of life (QOL), in particular, food has a strong connection between these two senses, and they have been treated mainly from chemical approach. Partly caused by these differences, the mechanism of the olfaction system had been controversial for a long time. Buck and Axel succeeded in cloning the odorant receptor that formed a multi-gene family in the rat olfactory epithelium in 1991¹, and received a Nobel Prize in Medicine and Physiology in 2004. Since their success in cloning, odorant receptor families have been found in rodents², mammals³, fish⁴, and insects⁵. With the attention of the researchers in this field and medical doctors of otolaryngology on this occasion, molecular biological research on odors has begun actively.

According to previous studies, it has been found that odorant molecules activate olfactory cells by binding to odorant receptors, and the information is converted to an electrical signal and transmitted to the olfactory bulb in the brain. It is also known that there are many types of odorant receptors and that one odor molecule reacts with multiple odorant receptors^6,7. For this reason, it is difficult to simply compare the patterns of reaction of odorant receptors because the response patterns of the odorant receptors are different for each odor. Furthermore, not all responses to human odorant receptors have been published for the protection of data rights. This correspondence has been investigated since around 1999⁸. Olfactory is receiving worldwide attention as Google is working on identifying odor from neural structures using neural network⁹ and technology that can detect odors on the ground was developed¹⁰. In addition, it is said that olfaction has a stronger connection with memory than other senses¹¹.

The human olfaction system is shown schematically in Fig. 1.¹². Smell is generally detected in the olfactory epithelium at the back of the nasal cavity. The olfactory epithelium is lined with millions of olfactory nerve cells. The olfactory epithelium is covered with mucous membranes, and olfactory nerve cells extend hairs called olfactory cilia in the mucous membrane. On the surface of the tip of the olfactory cilia, odor sensors called olfactory receptors are located. When odor molecules bind to olfactory receptors, the information is converted to electrical signals and transmitted to the olfactory bulb in the brain. Information transmitted to the olfactory bulb is perceived as the odor in the olfactory cortex. Because of this route through the limbic brain, which has a variety of functions including emotion, behavior, motivation, and long-term memory, smells are usually judged by not only the reaction of olfactory receptors but also through the experience and sensibility of individual persons.

In addition, as shown in Fig. 2., one odor molecule may react with multiple olfactory receptors instead of reacting with a single individual olfactory receptor⁶. It is already known that there are about 400 types of olfactory receptors in humans. In general, odorant molecules are volatile compounds with molecular weights of 30 to 300, and it is said that there are hundreds of thousands of odor molecules. And it is also known that the threshold values of each receptor to the concentration of odorant molecules differ for each molecule. By the combination of reactions with multiple receptors and differences in threshold values to molecules, we can discern a variety of smells.

Previous studies have used SVM (support vector machine) and GNN (Graph neural network) to infer olfactory receptor response patterns from olfactory receptor structures^19–21. Odor molecules with similar chemical structures are known to have similar odors. However, this does not apply to musk²² because the perception of the odor differs depending on the concentration of the odor molecules and the reaction patterns of olfactory receptors. Therefore, it is necessary to include information on the concentration at which odor molecules react with olfactory receptors in the network. In other words, it is thought that by focusing on the concentration at which odorants react with olfactory receptors and the pattern of reaction of olfactory receptors, odor identification can be performed close to the olfactory system.

The development of various sensors that respond to odor has been attempted these several decades. Semiconductor sensors utilize the phenomenon that the semiconductor comes into contact with specific odor molecules to change the electrical resistance and Quartz resonator sensors get the information of frequency changes when odor molecules are adsorbed on the surface of the quartz oscillator are typical examples ²⁴. Sensors of conductive polymer materials using the change in electrical resistance are another example. Moreover, sensors that utilize the expansion and contraction of sensitive membranes to which odor molecules are adsorbed have been developed. In the development of various types of odorant sensors, lots of efforts are being made to realize artificial olfaction or electric nose systems^25–27. However, current sensors have yet to reproduce the olfactory system of a living organism.

In this report, we proposed odorant receptor vectors consisting of 50% effective concentration (EC50) of odorant receptors for each odor molecule to examine various odors quantitatively. The EC50 is the value of the concentration at which an odorant molecule reacts with olfactory receptors at different concentrations, from the lowest maximum value to 50% of the maximum response. The EC50 is commonly used for drugs, antibodies, and other substances, and the concentration of each odorant molecule's response to olfactory receptors can be determined by using the EC50. Although odors had been evaluated qualitatively by human sensory tests, the similarity of odors could be evaluated quantitatively by using the inner product of odorant receptor vectors. In addition, the possibility of identifying odor molecules from the degree of activation of odorant receptors was studied by artificial neural networks.

The ability to identify odor molecules mechanically will lead to the development of artificial olfaction systems, which can be mounted on a robot, for example, to be useful for human health management and sniffing out food spoilage. Furthermore, it is expected to save lives at disaster sites by mounting it on a pet robot, etc. It has great future expectations.

Odorant receptor vectors

The number of odorant receptors in animals varies depending on species, and 396 receptors have been identified in humans by 2014²⁸. Given that 347 types of receptors were known around 2001^29,30, there is still a possibility that the number will increase in the future. The number of receptors is deeply related to the history of human beings spreading on the earth after being born on the earth in the eastern African continent, and there are also individual and regional differences in the number of receptors. These findings on olfactory receptors are said to be due to the significant improvement in the performance of the equipment for analyzing genomes and genes and the computational speed of computers.

Sensory evaluation tests have been used to characterize odors and determine the similarity of odors, but the odorant similarity can be expressed more clearly by quantifying each odor. Therefore, we propose a vector in which the number of olfactory receptors is defined as a dimension, and each element is a 50% effective concentration (described as EC50) of each receptor for odor molecules. This vector is called “the odorant receptor vector” in this paper. Previous studies have shown that one odor molecule binds to multiple olfactory receptors^8,31.

Therefore, the similarity of different odors can be expressed by taking the inner product of the odorant receptor vectors. Assuming that there are two different odors, the inner product is zero if the responding odorant receptors are completely different, while a finite non-zero number is calculated if some of the odorant receptors respond. The calculated number by the inner product of two different vectors means the similarity of odors.

Evaluation of similarity by odorant receptor vector

To evaluate the similarity of odors quantitatively, we performed several trials of calculation of the inner product of odorant receptor vectors using published literature data on EC50 at which odorant receptors reacted to odor molecules. At this time, 62 types of reactions between odorant molecules and odorant receptors were published in the literature³². The correlation among 62 types of odor molecules was studied using the inner product of the olfactory receptor vector for each odor molecule.

An example of the correlation coefficients among odorant molecules is shown in Figure 3. There have been published very few articles presenting row data of odorant molecules, and the molecules used in the sensory evaluation test in the literature were limited to 25 kinds as shown in the next section, then the correlation coefficients among 25 molecules are shown in Figure 3. 25 kinds of odorant molecules are arranged vertically and horizontally, and the inner product is expressed by the shade of color. The dark part indicates that the two odors highly resemble each other. In this way, the similarity of odor can be quantitatively evaluated using the inner product of the proposed odorant receptor vector. To confirm whether the inner products of the odorant receptor vector correctly express the similarity of odor, they were compared with the results by sensory evaluation tests conducted for 25 kinds of odorant molecules. Although the results of sensory evaluation tests sometimes vary widely among human subjects, the odor itself is exactly evaluated by individuals and the results by the examinee may be significant.

In the sensory evaluation test results were evaluated by 20 types of indexes for 200 types of odor molecules every 55 examinee³². In this test, each odor molecule was evaluated with a value from zero to 100 for each index. In this paper, 25 molecules were selected from 200 kinds of molecules which coincided with the prescribed 62 types of odorant molecules in the literature³¹. Then, the correlation coefficients among the evaluation of each index for 25 kinds of odorant molecules were calculated.

In another study, 12 panelists evaluated 48 odor molecules for the 41 odor descriptions given³³. 7 types of odor molecules have a common inner product of the olfactory receptor vector with this result. This is also considered to be a sensory evaluation.

Correlation coefficient

The correlation coefficient, r, is defined by the following Eq. (1).

where, s_xy is the covariance of x and y, s_x and s_y are the standard deviations of x and y, n is the total number of data (x, y), x_i and y_i are individual numerical values, and and are the respective mean values.

Artificial neural network

Research on artificial neural networks were started with a neuronal information transmission model proposed by W.S. McCulloch and W. Pitts in 1943³⁴. The application fields of the neural network were increasingly expanded by the proposal of various learning methods as an optimization problem of parameters in the network. With the background of deep learning technology proposed by G.E. Hinton et al. in 2006³⁵, the research on the application of artificial neural networks have been regarded as a core technology of artificial intelligence. The artificial neural networks can just make a mathematical mapping of input and output relations prepared as teaching signals, but they have a high ability in pattern recognition as nonlinear mappings. Several attempts to use artificial neural networks to identify odors have already been made by Nakamoto et al. since 1990³⁶. They proposed a conventional type of artificial neural network in which the outputs from multiple types of gas odor sensors were used as the input to the network, and the corresponding odors were the output of the neural network. And they concluded that the artificial neural network could make a successful nonlinear mapping for odor identification.

In this paper, quite new inputs of the artificial neural network for odor mapping were proposed. As the inputs to the network, EC50 of the odorant receptor vector was used, whereas the output was odor molecules. The conventional method of odor identification is based on the presence or absence of receptor activity^19,21, but this method is taken into account the activity concentration of olfactory receptors, allowing for more complex odor identification.

The artificial neural network had a normal three-layer hierarchical structure. The purpose of this study was to show that the correspondence between 63 types of odorant receptors and 62 types of odor molecules shown in the literature³¹ could be expressed by the artificial neural network. The number of units was 63 for the input layer and the hidden layer, and 62 for the output layer. The structure of the artificial neural network is shown in Figure 4. The mapping function was a conventional sigmoid function, and the learning algorithm was the normal back-propagation method.

Evaluation of similarity by odorant receptor vector

As an example, the correlation coefficients between cyclohexanone and each odorant molecule obtained from a sensory evaluation test were compared with the results from the normalized inner product value of the odorant receptor vector in Figure 5. The correlation between a certain target molecule and other odorant molecules based on the inner product obtained from the odorant receptor vector might express the similarity in chemical meaning, and several correlation coefficients of molecules showed good agreement with each other as shown in Figure 5. These highly correlated molecules may be regarded as the odorant ones with small individual differences and are easy to be identified. But if the correlation between the odors obtained by the two methods is significantly different, it is possible to judge the target odorant as the one difficult to distinguish. This may be caused by the pathway of information of odor in the brain, and the target odor is susceptible to memory by the experience of individuals.

Furthermore, the relationship between the normalized value of the inner product of the odorant receptor vector and the correlation coefficient between the odorant molecules in the sensory evaluation test was investigated. To evaluate the similarity of correlation distribution for each odorant molecule, the correlation coefficient derived by the two methods was calculated again for each of the 25 kinds of odorant molecules. The results are shown in Table 1 and Figure 6. Equation (1) is generally used in statistics, but when a continuous function is understood in vector space, the correlation coefficient between functions and the inner product between vectors have the same meaning. Therefore, by examining the correlation coefficient between the correlation distribution derived by the normalized inner product of odorant receptor vectors and that by the sensory evaluation test results, we can compare the target odor from the viewpoint of objective and subjective evaluation. When a large correlation is shown between them, it indicates that the target odor is easy to identify for many people, and in the case of a small correlation, it may be regarded as the odor with a large individual difference.

It is thought that it is possible to make foods more reproducible by adding flavors with odors that differ little among individuals. Focusing on Octanal, Hexanal, and Decanal, which are common in the two trials, we can see that they show similar trends. From this, it is possible to quantitatively evaluate the individual differences of each odor molecule by using the odorant receptor vector. By evaluating this individual difference, it is possible to understand the odor that people like, and it is thought to be useful when seasoning and adding fragrances.

Identifying odor by artificial neural network

The teaching signal was prepared for the neural network according to EC50 for 63 types of odorant receptors corresponding to 62 kinds of odors shown in the literature¹⁴. Regarding 63 types of mapping between EC50 of odorant receptors and odorant molecules as 1 cycle, a lot of cycles of training were conducted to optimize the parameters in the neural network. The average error transition of the output from the teaching signal is shown in Figure 7. The error decreased quickly by 1,000 cycles and turned to decrease slowly approaching zero, which meant that the pattern recognition of odors was performed.

There were two types of patterns in error transition; quick learning patterns, and slow and unrecognizable ones. For the patterns of quick learning, it was possible to identify odor molecules whose responses to odorant receptors differed greatly. Easily identifiable odor molecules were those molecules in which multiple receptors reacted at relatively wide concentrations, such as allylbenzene or caproic acid. On the other hand, it was difficult to identify those with similar responses to odorant receptors, such as hexyl acetate, octanal, etc. These molecules were said to make a reaction of a limited number of the same odorant receptors even if EC50 of odorant receptors were quite different. At present, EC50 as the input for the artificial neural network is not automatically obtained by odorant sensors, but it is considered that the development of new types of sensors capable of detecting EC50 in the future will enable such artificial intelligence to identify various odors.

Table 1. Correlation between the inner product of a vector and (A)sensory test results of 25 kinds of odorant molecules, (B) sensory test results of 7 kinds of odorant molecules

Odorant mole.	Correlation coef.	Odorant mole.	Correlation coef.	Odorant mole.	Correlation coef.
(-)-Carvone	0.514	Benzyl acetate	0.170	Heptanoic acid	0.731
2-Heptanone	0.050	Butyl formate	0.295	Hexanal	0.391
2-Hexanone	0.324	Coumarin	0.569	Hexyl acetate	0.349
2-Nonanone	0.189	Cyclohexanone	0.723	Octanal	0.505
2-Octanone	0.298	Decanal	0.646	Octanoic acid	0.562
3-Octanone	0.364	Decanoic acid	0.458	Propionic acid	0.748
Acetophenone	0.605	Dihydrojasmone	0.599	Allyl heptanoate	0.398
Benzene	0.626	Geraniol	0.473	Allyl phenylacetate	0.471
Benzophenone	0.502

Odorant mole.	Correlation coef.
2-Pentanone	0.579
Hexanal	0.374
Octanal	0.439
Decanal	0.404
Pentanoic acid	0.344
Carvone	0.344
Hexanoic acid	0.102

In this paper, we proposed a vector consisting of EC50 of odorant receptors and compared the inner product of the vectors with the results of a sensory evaluation test, using the published response of odorant receptors to odorant molecules at each concentration. It was shown that the correlation among each odorant molecule and also between the inner product of odorant receptor vectors and the result by sensory evaluation tests could be evaluated quantitatively. In addition, EC50 of each odorant receptor was arranged in the input layer of the artificial neural network, and it was shown that the odor could be identified by an artificial neural network using the odor molecule as the output layer. By using the concentration at which the odor molecules react with the receptors as a parameter, more complex odor identification becomes possible. Various new odorant sensors will be developed in the future. If they can be applied to the odorant receptor vector and the artificial neural network proposed in this paper, a machine with a discriminating function of odors will be developed for medical treatment, crisis management, crime prevention, etc.

ACKNOWLEDGMENT

The authors thank Professor Kazushige Touhara of University of Tokyo, and Professor Masayoshi Kobayashi of Mie University for their valuable suggestions.

Author contributions

KK developed Odorant receptor vectors and the original idea. Analyses of results by KK. SM was as major contributor in writing the manuscript. All authors have read and approved the manuscript.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that no competing interests.

AUTHOR DETAILS

¹Student of Mechanical Engineering, Yokohama National University, Kanagawa, Japan ²Emeritus Professor, Yokohama National University, Kanagawa, Japan

EC50, effective concentration

Buck, L. Axel, R. A Novel Multigene Family May Encode Odorant Receptors: A Molecular Basis for Odor Recognition. Cell. 1991; 65(1): 175–187. 1Student of Mechanical Engineering, Yokohama National University,2Emeritus Professor, Yokohama National University
Young, J. Trask, B. The sense of smell: genomics of vertebrate odorant receptors. Human Molecular Genetics. 2002; 11: 1153–1160.
Glusman G, Yanai I, Rubin I. Lancet, D. The Complete Human Olfactory Subgenome. Genome Res. 2001;11: 685–702.
Aparicio S, Chapman J, Stupka E, Putnam N, Chia J, Dehal P, Christoffels A, Rash S, Hoon Shawn, Smit A, Sollewijn Gelpke D M, Roach Jared, Oh Tania, Ho Y I, Wong M, Detter C, Verhoef F, Predki P, Tay A, Lucas S, Richardson P, Smith F S, Clark S M, Edwards J K Y, Doggett N, Zharkikh A, Tavtigian V S, Pruss D, Barnstead M, Evans C Baden H, Powell J, Glusman G, Rowen L, Hood L, Tan Y H, Elgar G, Hawkins T, Venkatesh B, Rokhsar D, Brenner S. Whole-Genome Shotgun Assembly and Analysis of the Genome of Fugu rubripes. Science, 2002;297:1301–1310.
Hill C, Fox N, Pitts J, Kent L, Tan P. Chrystal, A. M.; Cravchik, A.; Collins, H. F.; Robertson, M. M.; Zwiebel, J. L.; G Protein–Coupled Receptors in Anopheles gambiae. Science. 2002; 298: 176–178.
Touhara K. How Do We Sense Scents?: Molecular Basis for Odorant Disctimination by Olfactory Receptors(In Japanese). The Biophysical Society of Japan. 2004; 44: 26–31.
Yoshikawa E, Touhara K. Olfactory Receptor Function. Handbook of Olfaction and Gustation, edition no.3; Eds. Doty, L.R.; John Wiley & Sons, Inc.: New Jersey, 2015; pp109-121.
Malnic B, Hirono J, Sato T, Buck L, Combinatorial receptor codes for odors. Cell. 1999; 96: 713–723.
Sanchez-Lengeling B, Wei J N, Lee B K, Gerkin R C, Aspuru-Guzik A, Wiltschko A B. Machine Learning for Scent: Learning Generalizable Perceptual Representations of Small Molecules. arXiv. 2019. doi: https://doi.org/10.48550/arXiv.1910.10685
Yang Z, Sassa F, Hayashi K. Robot Equipped with a High-Speed LSPR Gas Sensor Module for Collecting Spatial Odor Information from On-Ground Invisible Odor Sources. ACS Sens. 2018; 3: 1174–1181.
Dahmani L, Patel R, Yang Y, Chakravarty M, Fellows L, Bohbot V. An intrinsic association between olfactory identification and spatial memory in humans. Nature Communications. 2018; 9: 4162.
The official website of the Nobel Prize, The Nobel Prize in Physiology or Medicine 2004. https://www.nobelprize.org/prizes/medicine/2004/press-release. Accessed 25 April 2022
Psolster J, Schieberle P. Structure – Odor Correlations in Homologous Series of Alkanethiols and Attempts To Predict Odor Thresholds by 3D-QSAR Studies. J. Agric. Food Chem. 2015; 63: 1419–1432.
Psolster J, Schieberle P. Structure – Odor Correlations in Homologous Series of Mercaptoalkanols. J. Agric. Food Chem. 2017; 65: 4329–4340.
Elsharif S A, Buettner A, Structure – Odor Relationship Study on Geraniol, Nerol, and Their Synthesized Oxygenated Derivatives. J. Agric. Food Chem. 2016; 66: 2324–2333.
Lorber K, Buettner A. Structure – Odor Relationships of (E)–3-Alkenoic Acids, (E)–3-Alken-1-ols, and (E)–3-Alkenals. J. Agric. Food Chem. 2015; 63: 6681–6688.
Lorber K, Zeh G, Regler J, Buettner A. Structure – Odor Relationships of (Z)–3-Alken-1-ols, (Z)–3-Alkenals, and (Z)–3-Alkenoic Acids. J. Agric. Food Chem. 2018; 66: 2334–2343.
Meijboom P W. Relationship Between Molecular Structure and Flavor Perceptibility of Aliphatic Aldehydes. J. American Oil Chem. Society. 1964; 41: 326–328.
Kowalewski J, Ray A. Predicting Human Olfactory Perception from Activities of Odorant Receptors. iScience. 2020; 23: 101361.
Ji H B, Seogjoo J J, Changbong H, Modular structure of human olfactory receptor codes reflects the bases of odor perception. bioRxiv, 2019. doi: https://doi.org/10.1101/525287
Keller A, Gerkin R C, Guan Y, Dhurandhar A, Turu G, Szalai B, Mainland J D, Ihara Y, Yu C W, Wolfinger R, Vens C, Schietgat L, Grave K D, Norel R. DREAM Olfaction Prediction Consortium; Stolovitzky, G.; Cecchi, G. A.; Vosshall, L. B.; Meyer, P.; Predicting human olfactory perception from chemical features of odor molecules. Science. 2017; 355: 820–826.
Chacko R, Jain D, Patwardhan M, Puri A, Karande S, Rai B. Data based predictive models for odor perception. Scientific Reports. 2020; 10: 17136.
Lavine B K, White C, Mirjankar N, Sundling C M, Breneman C M. Odor–Structure Relationship Studies of Tetralin and Indan Musks. Chem. Senses. 2012; 37: 723–736.
Ehara K. Trend of metal oxide semiconductor odor sensor (In Japanese). Society of Environmental Conservation Engineering. 1991; 20: 315–321.
Persaud K, Dodd G. Analysis of discrimination mechanisms in the mammalian olfactory system using a model nose. Nature. 1982; 299: 352–355
Romain C A, Nicolas J. Long term stability of metal oxide-based gas sensors for e-nose environmental applications: An overview. Sensors and Actuators B: Chemical. 2010; 2: 502–506
Ngo T H, Minami K, Imamura G, Shiba K, Yoshikawa G. Effects of Center Metals in Porphines on Nanomechanical Gas Sensing. sensors. 2018; 18: 1640.
Niimura Y, Matsui A, Touhara K. Extreme expansion of the olfactory receptor gene repertoire in African elephants and evolutionary dynamics of orthologous gene groups in 13 placental mammals. Genome Res. 2014; 24: 1485–1496.
Zozuya S, Echeverri F, Nguyen T. The human olfactory receptor repertoire. Genome Biology. 2001; 2: 0018.1.
Glusman G, Yanai I, Rubin I, Lancet D. The complete human olfactory subgenome. Genome Res. 2001; 11: 685
Saito H, Chi Q, Zhuang H, Matsunami H, Mainland J D. Odor Coding by a Mammalian Receptor Repertoire. Science Signaling. 2009; 2: ra9.
Keller A, Vosshall L B. Olfactory perception of chemically diverse molecules. BMC Neuroscience. 2016; 17: 55.
Guichard E, Barba C, Thomas-Danguin T, Tromelin A. Multivariate Statistical Analysis and Odor-Taste Network To Reveal Odor-Taste Associations. J. Agric. Food Chem. 2020; 68: 10318–10328
McCulloch S W, Pitts W. A logical calculus of the ideas immanent in nervous activity, The bulletin of mathematical biophysics. 1943; 5: 115–133
Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets. Neural Computation. 2006; 18: 1527–1554.
Nakamoto T, Fukunishi K, Moriizumi T. Identification capability of odor sensor using quartz-resonator array and neural-network pattern recognition. Sensors and Actuators B: Chemical. 1990; 1: 473–476

No competing interests reported.

Download PDF

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Odor Identification Quantitatively by Odorant Receptor Vectors

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Results And Discussion

Conclusions

Declarations

Abbreviations

References

Additional Declarations

Status:

Version 1