OdoriFy: A conglomerate of artificial intelligence–driven prediction engines for olfactory decoding

Abstract

The molecular mechanisms of olfaction, or the sense of smell, are relatively underexplored compared with other sensory systems, primarily because of its underlying molecular complexity and the limited availability of dedicated predictive computational tools. Odorant receptors (ORs) allow the detection and discrimination of a myriad of odorant molecules and therefore mediate the first step of the olfactory signaling cascade. To date, odorant (or agonist) information for the majority of these receptors is still unknown, limiting our understanding of their functional relevance in odor-induced behavioral responses. In this study, we introduce OdoriFy, a Web server featuring powerful deep neural network–based prediction engines. OdoriFy enables (1) identification of odorant molecules for wildtype or mutant human ORs (Odor Finder); (2) classification of user-provided chemicals as odorants/nonodorants (Odorant Predictor); (3) identification of responsive ORs for a query odorant (OR Finder); and (4) interaction validation using Odorant–OR Pair Analysis. In addition, OdoriFy provides the rationale behind every prediction it makes by leveraging explainable artificial intelligence. This module highlights the basis of the prediction of odorants/nonodorants at atomic resolution and for the ORs at amino acid levels. A key distinguishing feature of OdoriFy is that it is built on a comprehensive repertoire of manually curated information of human ORs with their known agonists and nonagonists, making it a highly interactive and resource-enriched Web server. Moreover, comparative analysis of OdoriFy predictions with an alternative structure-based ligand interaction method revealed comparable results. OdoriFy is available freely as a web service at https://odorify.ahujalab.iiitd.edu.in/olfy/. The molecular mechanisms of olfaction, or the sense of smell, are relatively underexplored compared with other sensory systems, primarily because of its underlying molecular complexity and the limited availability of dedicated predictive computational tools. Odorant receptors (ORs) allow the detection and discrimination of a myriad of odorant molecules and therefore mediate the first step of the olfactory signaling cascade. To date, odorant (or agonist) information for the majority of these receptors is still unknown, limiting our understanding of their functional relevance in odor-induced behavioral responses. In this study, we introduce OdoriFy, a Web server featuring powerful deep neural network–based prediction engines. OdoriFy enables (1) identification of odorant molecules for wildtype or mutant human ORs (Odor Finder); (2) classification of user-provided chemicals as odorants/nonodorants (Odorant Predictor); (3) identification of responsive ORs for a query odorant (OR Finder); and (4) interaction validation using Odorant–OR Pair Analysis. In addition, OdoriFy provides the rationale behind every prediction it makes by leveraging explainable artificial intelligence. This module highlights the basis of the prediction of odorants/nonodorants at atomic resolution and for the ORs at amino acid levels. A key distinguishing feature of OdoriFy is that it is built on a comprehensive repertoire of manually curated information of human ORs with their known agonists and nonagonists, making it a highly interactive and resource-enriched Web server. Moreover, comparative analysis of OdoriFy predictions with an alternative structure-based ligand interaction method revealed comparable results. OdoriFy is available freely as a web service at https://odorify.ahujalab.iiitd.edu.in/olfy/. The sense of smell allows an organism to sense its surroundings by recognizing and processing the information from diverse chemical clues present within the environment. These clues are largely composed of thousands of structurally diverse odorant molecules that mediate vital behavioral responses, such as social communication, identification and quality assessment of food (1Boesveldt S. Parma V. The importance of the olfactory system in human well-being, through nutrition and social behavior.Cell Tissue Res. 2021; 383: 559-567Crossref PubMed Scopus (9) Google Scholar), and the recognition of prey and predators (2Hughes N.K. Price C.J. Banks P.B. Predators are attracted to the olfactory signals of prey.PLoS One. 2010; 5e13114Crossref PubMed Scopus (68) Google Scholar, 3Hussain A. Saraiva L.R. Ferrero D.M. Ahuja G. Krishna V.S. Liberles S.D. Korsching S.I. High-affinity olfactory receptor for the death-associated odor cadaverine.Proc. Natl. Acad. Sci. U. S. A. 2013; 110: 19579-19584Crossref PubMed Scopus (118) Google Scholar). Olfactory receptors, the functional units of an olfactory system, are the key drivers of odor perception as they contribute to the first critical step of odorant detection (3Hussain A. Saraiva L.R. Ferrero D.M. Ahuja G. Krishna V.S. Liberles S.D. Korsching S.I. High-affinity olfactory receptor for the death-associated odor cadaverine.Proc. Natl. Acad. Sci. U. S. A. 2013; 110: 19579-19584Crossref PubMed Scopus (118) Google Scholar, 4Sarafoleanu C. Mella C. Georgescu M. Perederco C. The importance of the olfactory sense in the human behavior and evolution.J. Med. Life. 2009; 2: 196-198PubMed Google Scholar, 5Raman B. Ito I. Stopfer M. Bilateral olfaction: Two is better than one for navigation.Genome Biol. 2008; 9: 212Crossref PubMed Scopus (9) Google Scholar, 6Gaillard I. Rouquier S. Giorgi D. Olfactory receptors.Cell. Mol. Life Sci. 2004; 61: 456-469Crossref PubMed Scopus (107) Google Scholar, 7Ahuja G. Korsching S. Zebrafish olfactory receptor ORA1 recognizes a putative reproductive pheromone.Commun. Biol. Scopus Google Scholar). In these of receptors are primarily in the olfactory of the olfactory to other the human olfactory such as and olfactory sensory and olfactory A. S. S. Ahuja D. Ahuja G. The basis of the of smell in Scopus Google Scholar, S. S. analysis of olfactory and in PubMed Scopus Google Scholar, of human olfactory a The Res. PubMed Scopus Google Scholar, The of the olfactory in PubMed Scopus Google Scholar). the olfactory sensory functional relevance they the receptors that are for the recognition and discrimination of odorant olfactory sensory the that a functional receptor S. S. analysis of olfactory and in PubMed Scopus Google Scholar, Olfactory sensory olfactory receptors Biol. PubMed Scopus Google Scholar, S. receptor in the olfactory 2004; PubMed Scopus Google Scholar). The from the olfactory sensory the receptor to a of the olfactory C. A. M. an olfactory sensory PubMed Scopus Google Scholar, G. S. V. V. D. Korsching S.I. a of olfactory sensory PubMed Scopus Google Scholar, G. I. M. Korsching S.I. Zebrafish to a 2013; PubMed Scopus Google Scholar, I. A. of olfactory sensory by odorant receptors and the 2004; PubMed Scopus Google Scholar, A. D. of in the human olfactory for odor One. 2008; PubMed Scopus Google Scholar). The olfactory system is composed of of These odorant receptors receptors A. S. S. Ahuja D. Ahuja G. The basis of the of smell in Scopus Google Scholar), and receptors S. S. analysis of olfactory and in PubMed Scopus Google Scholar), receptors of human olfactory a The Res. PubMed Scopus Google Scholar), and The of the olfactory in PubMed Scopus Google Scholar), and the A receptors Olfactory sensory olfactory receptors Biol. PubMed Scopus Google Scholar). in for the functional and I. Rouquier S. Giorgi D. Olfactory receptors.Cell. Mol. Life Sci. 2004; 61: 456-469Crossref PubMed Scopus (107) Google Scholar, G. S. D. The human olfactory to and PubMed Scopus Google Scholar, M. of olfactory receptor in the human Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar), other a of functional receptors S. receptor in the olfactory 2004; PubMed Scopus Google Scholar, C. A. M. an olfactory sensory PubMed Scopus Google Scholar). the human functional and receptor M. of receptor in and of a of receptors in the olfactory Biol. PubMed Scopus Google Scholar, of olfactory and the Biol. 2008; PubMed Scopus Google Scholar). receptors and receptors M. of receptor in and of a of receptors in the olfactory Biol. PubMed Scopus Google Scholar, of olfactory and the Biol. 2008; PubMed Scopus Google Scholar, M. A. I. to olfactory of Tissue Res. 2021; 383: PubMed Scopus (4) Google Scholar). Moreover, functional receptors to their functional importance in detection is highly V. S. M. G. G. V. M. receptors and their putative in Med. Sci. Google Scholar). This is by the that functional that a in receptors B. M. G. D. M. the of the human a Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). A of human olfactory revealed the of or as as A L.R. M. C.J. M. C. M. M. M. C. S. of olfactory an on food odor detection in PubMed Scopus Google Scholar). the the of that olfactory receptors a to the identification of a myriad of structurally odorant molecules (3Hussain A. Saraiva L.R. Ferrero D.M. Ahuja G. Krishna V.S. Liberles S.D. Korsching S.I. High-affinity olfactory receptor for the death-associated odor cadaverine.Proc. Natl. Acad. Sci. U. S. A. 2013; 110: 19579-19584Crossref PubMed Scopus (118) Google Scholar, 4Sarafoleanu C. Mella C. Georgescu M. Perederco C. The importance of the olfactory sense in the human behavior and evolution.J. Med. Life. 2009; 2: 196-198PubMed Google Scholar, 5Raman B. Ito I. Stopfer M. Bilateral olfaction: Two is better than one for navigation.Genome Biol. 2008; 9: 212Crossref PubMed Scopus (9) Google Scholar, 6Gaillard I. Rouquier S. Giorgi D. Olfactory receptors.Cell. Mol. Life Sci. 2004; 61: 456-469Crossref PubMed Scopus (107) Google Scholar, 7Ahuja G. Korsching S. Zebrafish olfactory receptor ORA1 recognizes a putative reproductive pheromone.Commun. Biol. Scopus Google Scholar). In to other because of its of functional the of ORs a of and identification of their agonists and B. The human olfactory receptor Natl. Acad. Sci. U. S. A. 2004; PubMed Scopus Google Scholar). we the agonists for chemicals that are known to and an to chemicals that to a functional in the interaction In the chemical an and an odor we the for to other ORs primarily on the of the olfactory sensory of the olfactory Olfactory on the odorant or to its the interaction an odorant at the of an in the of from the U. D. Olfactory of Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). These to the of U. D. mediate olfactory PubMed Scopus Google Scholar). the to the of U. D. mediate olfactory PubMed Scopus Google Scholar, A in olfactory receptor PubMed Scopus Google Scholar). the of the because of the of This by the of because of M. S. A. A. The of the in olfactory PubMed Scopus Google Scholar, D. D. mechanisms in olfactory receptor PubMed Scopus Google Scholar). the with in functional revealed of and of these ORs the S. A. V. A. S. D. Ahuja G. of of olfactory receptors with and Biol. PubMed Scopus Google Scholar, S. B. of the odorant receptor in human Biol. PubMed Scopus Google Scholar, A. I. D. odorant receptors, and receptors, are for of Biol. PubMed Scopus Google Scholar, I. of olfactory and PubMed Scopus Google Scholar). the in M. I. D. of olfactory receptor PubMed Scopus Google Scholar). A of ORs in the and the than ORs C. S. S. G. of olfactory receptors by deep One. 2013; PubMed Scopus Google Scholar, S. A. M. S. D. Ahuja G. and for olfactory PubMed Scopus (3) Google Scholar). in such as to and S. A. V. A. S. D. Ahuja G. of of olfactory receptors with and Biol. PubMed Scopus Google Scholar, I. of olfactory and PubMed Scopus Google Scholar, S. A. M. S. D. Ahuja G. and for olfactory PubMed Scopus (3) Google Scholar, S. Olfactory receptors in Google Scholar, A. M. A. A. S. N.K. S. M. M. V. D. A for and for 2021; Scopus (1) Google Scholar). a as a to S. B. of the odorant receptor in human Biol. PubMed Scopus Google Scholar, I. of olfactory and PubMed Scopus Google Scholar, of olfactory receptors in human and PubMed Scopus Google Scholar, S. M. D. A. B. M. A. G. Odorant receptor and PubMed Scopus Google Scholar). The functional of by acid in of the an and of G. I. A. S. S. D. of and by olfactory receptor One. PubMed Scopus Google Scholar), the or of A analysis revealed the of ORs in of at the resolution S. A. V. A. S. D. Ahuja G. of of olfactory receptors with and Biol. PubMed Scopus Google Scholar). in the as as other receptor is by the importance of the of I. of olfactory and PubMed Scopus Google Scholar). ORs a highly of functional are by a receptor 2009; PubMed Scopus Google Scholar, C. A. than olfactory PubMed Scopus Google Scholar, C.J. the of olfactory Scopus Google Scholar). In to a of and structurally diverse chemical with that a to their receptors G. M. M. A. C. C. M. M. M. provides Biol. PubMed Scopus Google Scholar, A. C. M. on odor it Biol. PubMed Google Scholar). because of their chemical are still to the chemical basis of odorant molecules C. A. for Scholar). understanding that functional perception for are with a smell or a smell, a smell S. odor classification through a PubMed Scopus Google Scholar). with the of odorant it that the underlying olfactory mechanisms are than In in the of in the perception of an odorant or A. M. A. A. S. N.K. S. M. M. V. D. A for and for 2021; Scopus (1) Google Scholar, M. S. A for understanding in PubMed Scopus Google Scholar, A. I. G. A. or a for from chemical PubMed Scopus Google Scholar, C. S. of odorant receptors are from chemical 9: PubMed Scopus Google Scholar, C. and in to odorant Mol. Biol. PubMed Scopus Google Scholar). the of such is still This primarily because of the of and to and of interaction To this we introduce OdoriFy, a comprehensive Web server for deep neural network–based prediction of human OdoriFy is built on a comprehensive repertoire of manually curated olfactory and interaction making it one of the curated to In OdoriFy prediction that Odorant and Odorant–OR Pair that allow prediction of odorant identification of responsive ORs for the and prediction of putative for wildtype or mutant In to OdoriFy of explainable artificial that the of key of the or ORs at the atomic or amino acid OdoriFy is an Web server with deep neural network–based prediction with explainable artificial is with the of a to chemical in the of OdoriFy is of agonists for human ORs (Odor prediction of odorant molecules (Odorant identification of responsive ORs for the chemicals (OR and Odorant–OR Pair Analysis. The provides with a highly and agonists for their of or the of odorant in the In the of Odorant the query chemicals in the of the or by in the To in their query chemicals the OdoriFy provides a to the Web server D.M. to an PubMed Scopus Google Scholar). Odorant that information the chemical of an odorant or its prediction and the artificial The prediction chemicals as in the provides that and the is as compared with the The of prediction is the as that of Odorant it provides the of the ORs the chemicals and information on the amino of the allows to the for the wildtype or allows to the of their ORs of using the or the the of to The prediction is to that of the prediction engines. for the Odorant–OR Pair the with the query and OdoriFy the interaction with the other information as for the other engines. OdoriFy as a by using the the prediction allow to their the prediction in OdoriFy are at the by The first is to the Odorant the one the prediction engines. these are built using deep neural G. for and PubMed Scopus Google (Odorant and on (OR and Odorant–OR Pair using to we first built the Odorant by using the of the G. for and PubMed Scopus Google Moreover, we its for the The in OdoriFy is built on In the a and on the amino acid and of agonists and is to the with a of and the is to to this the of and for V. M. B. A. C. S. A and for for at the amino acid and atomic levels. In the of Odorant we and and a of known odorant and molecules from the or for odorant and molecules and to the chemical of and in the we using the A. A for 2008; PubMed Scopus Google Scholar). the of and molecules in the of a these as we analysis and the chemical of the and in a revealed a of chemical these of the and at the functional revealed a relatively of functional such as and the are in with the The recognition of chemical by The of at the functional Scopus Google Scholar, S. M. of functional on of in the olfactory PubMed Scopus Google and for the we G. for and PubMed Scopus Google Scholar), a computational for and G. for and PubMed Scopus Google as and using the method and the predictive to The Odorant the of on the the of the to the of such as and a and the module of G. for and PubMed Scopus Google provides the putative of the classification at the atomic A and and and To the of deep neural in the we first a of information known for the ORs from the and to the quality of the we the and our by first through the Odorant to the odorant molecules for the analysis revealed in the with comparable for the functional the agonists and of the agonists or in our functional and the we built the prediction using deep neural and and and the by revealed the and of and Moreover, the of to for the and to the we an to the of and an with at the on and its on the The an of on the the of the to the Moreover, other for assessment within the and we this as a for the and Odorant–OR Pair prediction by leveraging one of the and highly of known information human interaction with a deep we built a prediction that allows of of the key distinguishing of the OdoriFy Web server is that it allows the of for prediction and validation of the In the of it of the odorant as an and a of with their interaction and the the of interaction the odorant and the Moreover, it highlights the of the odorant featuring their relevance in the making by the and A and A is for the amino acid with a that its importance in making and and A of in the prediction of chemicals as it in the and Moreover, it a of odorant with their interaction to the at the atomic for the odorant molecules and A and and amino acid for the ORs and and to the OdoriFy allows to a of or and This is in Odorant–OR Pair a prediction in the OdoriFy Web server. OdoriFy a of artificial prediction and allows of alternative to the for interaction is the ligand interaction analysis by molecular To date, is of present in the G. The Res. PubMed Scopus Google Scholar). an alternative by the recognition by the molecular is a for the prediction of and is the and putative odorant To the OdoriFy for are we a comparative analysis with one of the human we built a of A. M. A. A. S. N.K. S. M. M. V. D. A for and for 2021; Scopus (1) Google using the recognition method using and it using the molecular of of S. B. molecular through from to Scopus Google To OdoriFy we and agonists and nonagonists, and and with receptor 2009; PubMed Scopus Google within the of A and the and we known agonists and of a using and predictive interaction using of agonists and nonagonists, of these alternative we in the of the agonists or the of the structure-based method OdoriFy prediction as compared with the tools. To we an validation from our curated this validation as an query on other prediction that G. A web server for olfactory PubMed Scopus Google and A. I. neural for and their olfactory Biol. Google Scholar). the comparative analysis for the because of the functional with A. I. neural for and their olfactory Biol. Google and G. A web server for olfactory PubMed Scopus Google our analysis revealed that OdoriFy these alternative with it comparable with G. A web server for olfactory PubMed Scopus Google with In this we introduce OdoriFy, a comprehensive Web server with prediction that Odorant and Odorant–OR Pair Analysis. These prediction allow validation of odorant prediction of responsive ORs for prediction of putative for and of the prediction of explainable artificial the to the underlying for the at the atomic or and amino acid in the to prediction for as as for the classification G. A web server for olfactory PubMed Scopus Google Scholar, A. I. neural for and their olfactory Biol. Google Scholar). G. A web server for olfactory PubMed Scopus Google as as A. I. neural for and their olfactory Biol. Google manually molecular for as as for OdoriFy Moreover, the of explainable artificial is to and a of are the key of the and of The module for the query this is in the of G. A web server for olfactory PubMed Scopus Google and in the of A. I. neural for and their olfactory Biol. Google Scholar). OdoriFy is a Web server and and in to the OdoriFy is the are on of using deep OdoriFy G. A web server for olfactory PubMed Scopus Google and A. I. neural for and their olfactory Biol. Google Scholar), OdoriFy the chemical of the olfactory system it is built on human OdoriFy to and and therefore with a OdoriFy with such as for that OdoriFy the computational in the of to the for ORs provides a to for the in interaction as to the and Moreover, OdoriFy allow of the molecular basis of that to the amino acid to that of odorant a to These with the key amino of the that contribute to the and therefore the for the in The that the of the Odorant to the of odorant that to the of that and this to the of key of the using feature our understanding of odorant and therefore the of C.J. the of olfactory Scopus Google Scholar). OdoriFy server and its to the odorant and its odorant to odorant C. A. than olfactory PubMed Scopus Google Scholar, M. the of odor PubMed Scopus Google Scholar, The of olfactory that is still PubMed Scopus Google Scholar). OdoriFy the available to the with their receptors and A. A. G. B. C. human olfactory perception from chemical of odor PubMed Scopus Google Scholar). In to OdoriFy to on the of The majority of the human ORs are and therefore contribute to odorant detection M. of olfactory receptor in the human Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar, of olfactory receptor in and Biol. 2008; PubMed Scopus Google Scholar). to the for these such an other computational to and functional in is that the odor perception is by an odorant to its receptor of other ORs by the as a to odor C. M. G. M. Odorant receptor is to odor Biol. PubMed Scopus Google Scholar, M. Olfactory receptor 2004; PubMed Scopus Google Scholar). OdoriFy the of known a of to the molecular basis of using the or Moreover, OdoriFy contribute to the identification of for the A the that with the I. S. A of an olfactory PubMed Scopus Google Scholar). This that the ORs by the prediction to the ORs to that of user-provided I. S. A of an olfactory PubMed Scopus Google Scholar). and the underlying deep of OdoriFy other of such as in G. A. using deep 2021; 61: PubMed Scopus Google Scholar). a of artificial allows of or from the as and therefore the of classification or prediction on the are an and of A deep neural is the of that are to other The of information or signals and therefore to a that with The of deep a in A. G. and deep in The prediction of for 2013; PubMed Scopus Google Scholar, to artificial PubMed Scopus Google Scholar). The present deep in and therefore allows classification of chemicals and and identification of putative ORs for the query or The for OdoriFy prediction Odorant and Odorant–OR Pair manually from by PubMed the of largely on the for the we to the the and its in In the first interaction information to the information their known agonists and nonagonists, from and of the and the of the manually and In the the information and for by a In the the manually by the A of and for the for the Odorant classification allows of odorant in To the and we the and molecules from and To its we the to the and we a of and that we for the deep these chemical in the of a we a to we the D.M. to an PubMed Scopus Google for the of chemicals The of OdoriFy from the validation of odorant in query chemicals to the identification of their The of the Odorant module is on G. for and PubMed Scopus Google Scholar). with the for the The as the of the key of is that it allows of the for and the of with In the of Odorant the built using the an in G. for and PubMed Scopus Google Scholar). The of the at the using the relevance method S. A. G. for by relevance One. PubMed Scopus Google Scholar). relevance method is an in with neural to and allow the of relevance from the to the this the To and the of the prediction and we an in the of and by and its the of using the such as the the of and is as the of predictions by the as a of to an of in a as the These by the of that for provides information on the of predictions that are the provides information on the of the that are is the of and The provides information the the The the of the is the of the underlying to the the of the for the of The and Odorant–OR Pair built using a deep A. S. A. G. A. A. M. Scholar). a at the that of these prediction engines. The for with the of their and the and the of the we a of and as for the Odorant the first through Odorant to from the the majority of the of and of in the for odorant and therefore for the In the of the Odorant–OR Pair prediction the and The for the for the the interaction a of nonagonists, as compared with To we the using from the to from The and their agonists and from the of a of using the of the receptor by and one and of for the Two on the of and receptor these through the of the and their This by with and as the to using revealed the for and are and and to the for our on and validation by the responsive ORs for the query of the to the and their receptors, we our by this on a of are structurally to the This using the for and the molecular using the limited of we their with the query by In to this we a we the of the query with receptor available in our manually curated This step the of the with human ORs and the ORs with the information their and is and comprehensive than prediction G. with explainable artificial 2: Scopus Google Scholar, M. A. for Scholar). the prediction on the mutant and wildtype in the of the prediction for with wildtype receptors human The of the provides a of ORs with their for the user-provided In the prediction enables the identification of odorant molecules for wildtype or mutant human we the for our the underlying is the as the one we for the Odorant–OR Pair Analysis. of prediction are and to an module for prediction to the of a module the and with of and The module is built on the a of the importance of the to the prediction V. M. B. A. C. S. A and for Scholar, A. S. A. G. A. A. M. Scholar). in the and analysis and we the that molecules and to their relevance in the OdoriFy is built on the the with and from an to the Web such as of and to an of on the Web server and of processing an on the server the on the a is of the is using on the prediction Odorant and the of a with in is the of of the first the and the other the receptor Odorant–OR Pair on the other an of in its the on the and the to the with the from the The are using a and using an is to the that their query as as OdoriFy is an with prediction engines. is as a Web server Moreover, the of OdoriFy from The of the to the is available at This The that they of with the of this the of of for with the computational the of the Ahuja for their at of this G. A. G. D. A. V. S. S. S. V. and M. A. M. G. A. G. D. and A. M. and G. A. G. A. G. A. The Ahuja is by the a of the of of of from the and and an from of The is by the from the of

Topics & Concepts