D1210–D1216 Nucleic Acids Research, 2018, Vol. 46, Database issue Published online 20 October 2017 doi: 10.1093/nar/gkx957 FlavorDB: a database of flavor molecules Neelansh Garg1,2,†, Apuroop Sethupathy1,3,†, Rudraksh Tuwani1,4,†, Rakhi NK5,†, Shubham Dokania1,6,†,ArvindIyer1,†, Ayushi Gupta1,†, Shubhra Agrawal1,†, Navjot Singh1,6,†, Shubham Shukla1,7,†, Kriti Kathuria1,8,†, Rahul Badhwar5, Rakesh Kanji5, Anupam Jain5, Avneet Kaur1, Rashmi Nagpal1 and Ganesh Bagler1,* 1Center for Computational Biology, Indraprastha Institute of Information Technology (IIIT-Delhi), New Delhi, India, 2University School of Information, Communication and Technology, Guru Gobind Singh Indraprastha University, New Delhi, India, 3Ashoka University, Sonepat, Haryana, India, 4Sri Venkateswara College, Delhi University, New Delhi, India, 5Department of Bioscience and Bioengineering, Indian Institute of Technology Jodhpur, Jodhpur, India, 6Delhi Technological University, New Delhi, India, 7Northern India Engineering College, Guru Gobind Singh Indraprastha University, New Delhi and 8Maharaja Agrasen College, Delhi University, New Delhi, India Received August 15, 2017; Revised September 18, 2017; Editorial Decision October 05, 2017; Accepted October 06, 2017 ABSTRACT INTRODUCTION Flavor is an expression of olfactory and gusta- Flavor is a complex, multi-sensory human experience with tory sensations experienced through a multitude of a rich evolutionary history (1). Molecules form the chemi- chemical processes triggered by molecules. Beyond cal basis of flavor expressed primarily via gustatory and ol- their key role in defining taste and smell, flavor factory mechanisms. The perception of flavor arises from molecules also regulate metabolic processes with interaction of flavor molecules with the biological machin- ery and could be perceived as an emergent property of a consequences to health. Such molecules present complex biochemical system. While some components of in natural sources have been an integral part of this puzzle have been unearthed, a holistic view of this phe- human history with limited success in attempts nomenon still eludes us (2–5). Taking a data-centric ap- to create synthetic alternatives. Given their utility proach can provide a systems perspective of flavor sensation in various spheres of life such as food and fra- by offering ways to discern its key features. grances, it is valuable to have a repository of flavor Flavors derived from natural sources have shaped culi- molecules, their natural sources, physicochemical nary habits throughout human history. Analogous to vari- properties, and sensory responses. FlavorDB (http: ations in regional languages, cultures have evolved varia- //cosylab.iiitd.edu.in/flavordb) comprises of 25,595 tions in the way they cook. Traditional recipe compositions flavor molecules representing an array of tastes and encode ingredient combinations that are not only palat- odors. Among these 2254 molecules are associated able but appetizing. Heuristic associations between molecu- lar properties and perception of flavors provide indications with 936 natural ingredients belonging to 34 cate- towards its chemical basis (1). For example, combinations gories. The dynamic, user-friendly interface of the re- of aliphatic esters play a major role in many fruit flavors. source facilitates exploration of flavor molecules for Ketones are known to impart metallic flavors in oxidized divergent applications: finding molecules matching butter, and monoterpenoids provide the characteristic fla- a desired flavor or structure; exploring molecules of vors of many herbs and spices. However, such knowledge an ingredient; discovering novel food pairings; find- remains largely unstructured and incomprehensive. ing the molecular essence of food ingredients; as- FlavorDB was created with the aim of integrating mul- sociating chemical features with a flavor and more. tidimensional aspects of flavor molecules and representing Data-driven studies based on FlavorDB can pave the their molecular features, flavor profiles and details of nat- way for an improved understanding of flavor mecha- ural source (Figure 1). FooDB, one of the efforts in simi- nisms. lar direction, compiles molecules from food ingredients; al- though its focus is not on chemical basis of flavor or fla- vor pairing (http://foodb.ca). Flavornet is another resource, which provides a list of flavor molecules and their odor *To whom correspondence should be addressed. Tel: +91 11 26907443; Email: [email protected]; [email protected] †These authors contributed equally to this work as first authors. Present address: Ganesh Bagler, Center for Computational Biology, Indraprastha Institute of Information Technology (IIIT-Delhi), New Delhi 110020, India. C The Author(s) 2017. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected] Nucleic Acids Research, 2018, Vol. 46, Database issue D1211 Figure 1. FlavorDB is a seamless amalgamation of ‘entity space’ and ‘flavor space’. The resource provides a comprehensive dataset along with auser- friendly interface and interlinked search engines for exploring the flavor universe. profiles, but does not furnish information of their natural dients, flavor molecules and their flavor profile, and chem- sources (6). Other attempts in this direction have focused on ical descriptors including functional groups and physico- compilation of data specific to aspects of flavors: tastes such chemical properties. Through an extensive repertoire of in- as bitter (BitterDB) and sweet (SuperSweet), and volatile gredients and their constituent flavor molecules, FlavorDB compounds of scents (SuperScent) (7–9). Certain others also provides a tool for experimenting with flavor pairing. have targeted nutritional factors (NutriChem), polyphenols Thus, it offers an integrative platform for exploring the fla- (Phenol-Explorer) and the medicinal value of food (10–13). vor space with the help of a feature-rich visual interface. Among other sources, FlavorDB collates information FlavorDB combines different dimensions of flavor con- from FooDB, Flavornet, SuperSweet and BitterDB to cre- stituting the ‘entity space’ and ‘flavor space’ (Figure 1). The ate a comprehensive repository of flavor molecules, fla- former incorporates facets of ingredients which are enti- vor profiles, physicochemical properties and natural sources ties from natural sources often used in food, whereas the (Section S1, Supplementary Data). Compared to FooDB latter represents molecules responsible for flavor sensation which has 2816 flavor compounds, FlavorDB covers 25,595 and their descriptors. By bringing relevant information un- flavor molecules compiled from Fenaroli’s Handbook of der a single umbrella, FlavorDB provides a comprehensive Flavor Ingredients and literature survey in addition to in- dataset backed by a user-friendly interface, creative visual- tegrating data from all the above-mentioned sources. Fla- izations, and interlinked search engines for exploring fea- vorDB spans across 34 ingredient categories covering 936 tures that contribute to the sensation of flavor. Thus, Fla- ingredients of which 190 are unique. One of the features vorDB paves the way for an improved understanding of fla- which sets FlavorDB apart from similar resources is that it vor perception arising out of complex interplay of flavor presents information in a hierarchy of food category, ingre- compounds with biological systems and allied applications. D1212 Nucleic Acids Research, 2018, Vol. 46, Database issue DATABASE OVERVIEW ingredients were used to query PubMed to obtain articles that reported their flavor molecules. Flavor molecules as- FlavorDB is a resource with extensive coverage of 25,595 sociated with entities/ingredients were thus curated from flavor molecules (Figure 1). Among molecules listed in existing sources (15–17) (FooDB; http://foodb.ca,arXiv the database, 2254 have been reported to be found in 936 preprint arXiv:1502.03815, 2015) and compiled manually. natural entities/ingredients. These natural ingredients have Molecules from Flavornet, BitterDB and SuperSweet were further been classified into 34 categories, and mapped to further included along with their flavor profiles6 ( ,7,9). Ad- 527 distinct natural sources. An additional 13,869 com- ditionally, information for 33 taste receptors (Sweet, Bitter, pounds were identified as synthetic. For the remaining 9472 Sour, and Umami) and 1068 odor receptors is also available molecules, no specific source could be ascertained. The fea- in FlavorDB. For each receptor, we provide its Uniprot ID, tures provided as part of the detailed molecular and flavor name, involvement in taste, and Uniprot link (18). profiles of these compounds have an impact on their taste The chemical identifiers of molecules were obtained and odor through gustatory and olfactory sensory mecha- from various sources (FooDB; http://foodb.ca)(6,7,9,15,16) nisms. (Also see Section S2, Supplementary Data), and were stan- FlavorDB offers a user-friendly interface for querying dardized to procure their CAS (Chemical Abstract Ser- and browsing flavor molecules, entities/ingredients, natu- vice) numbers. CAS numbers were