Contents lists available at ScienceDirect Food Research International journal homepage: www.elsevier.com/locate/foodres Combinations of graph invariants and attributes of simplified molecular input-line entry system (SMILES) to build up models for sweetness P.G.R. Achary a, ⁎ , A.P. Toropova b , A.A. Toropov b a Department of Chemistry, Institute of Technical Education and Research (ITER), Siksha ‘O’ Anusandhan deemed to be University, Bhubaneswar, Odisha 751030, India b Department of Environmental Health Science, Laboratory of Environmental Chemistry and Toxicology, Istituto di RicercheFarmacologiche Mario Negri IRCCS, Via La Masa 19, 20156 Milan, Italy ARTICLE INFO Keywords: Sweetness potential QSAR OECD principles Monte Carlo method Index of Idelaity of correlation CORAL software ABSTRACT The quantitative structure – activity relationships (QSARs) for sweetness value (log S) were built with a dataset of 315 molecules; following a novel criterion of ‘Index of Ideality of Correlation(IIC)’ This criterion of IIC is available in the latest version of the CORAL software (www.insilico.eu/coral). The descriptor used in the model building for log S is a hybrid optimal descriptor; obtained by combining the two descriptors: (i) molecular graph based descriptor derived from correlation weights of molecular features and (ii) descriptor derived from the simplified molecular input-line entry system (SMILES) code of sweetener molecule. The data set of 315 mole- cules was divided into four random splits. The four QSAR models which were build for log S using the criterion of IIC were compared with four similar models built “traditional protocol” described elsewhere. The comparison revealed that the models built using IIc were better with statistical performance. 1. Introduction Today, type 2 diabetes, cardiovascular diseases and obesity are becoming common diseases globally (Lustig, Schmidt, & Brindis, 2012). The well accepted reason for the increase in the number of patients in the above diseases is the change in lifestyle and the consumption of high calorie food. One of the major steps to minimize the risk and to decrease the calorie intake a good number of low-calorie sweeteners are coming in the market and becoming a part of the life style. The growing market of low-calorie sweeteners is not only due to the above patients but also huge consumption of people (non-patients) working in the philosophy of ‘Prevention is better than cure’. The sweeteners synthe- sized chemically or extracted from natural products are subject of dis- cussion as far as health and safety is concerned (Bassoli, Borgonovo, & Morini, 2011; Behrens, Meyerhof, Hellfritsch, & Hofmann, 2011; DuBois & Prakash, 2012). Literature review on ‘SweetenersDB’ reveal that 316 sweeteners have the sweetness value(S) in between 0.200 and 2.25 × 10 5 , the definition of relative sweetness is presented in the data section. Due to wide distribution of the sweetness values it is better expressed as logarithm of sweetness value(log S).The sweetness potential is an im- portant data for chemistry, psychology, pharmacology, biochemistry, and the food industry (Chéron, Casciuc, Golebiowski, Antonczak, & Fiorucci, 2017).The definition of SP is not fast and cheap. Hence, the search for computational methods to estimate SP is actually a vital task. Quantitative structure – activity relationships (QSARs) are a tool to build up predictive models of SP for substances, which were not studied experimentally (Goel, Gajula, Gupta, & Rai, 2018; Zhong, Chong, Nie, Yan, & Yuan, 2013).Among the various tools to design predictive models (QSAR) for different endpoints of various substances CORAL software is known for its uniqueness (Amata et al., 2017; Choi, Trinh, Yoon, Kim, & Byun, 2019; Fioressi, Bacelo, Rojas, Aranda, & Duchowicz, 2019; Ghaedi, 2015; Islam & Pillay, 2016; Rescifina et al., 2017; Toropov, Achary, & Toropova, 2016; Toropov et al., 2019; Toropov & Toropova, 2018; Toropova & Toropov, 2019; Toropova, Toropov, Begum, & Achary, 2018; Velázquez-Libera, Caballero, Toropova, & Toropov, 2019).Thus, with a motivation that the CORAL software can also give good models for the sweetness value (log S); different possibilities, functionalities uniqueness available in CORAL was explored in this communication. Recently, a good number of papers used ‘Index of Ideality of Correlation(IIC)’ as a novel criterion to build best predictive QSAR models (Toropov, Carbó-Dorca, & Toropova, 2018; Toropov & Toropova, 2017, 2018; Toropova & Toropov, 2017b, 2018). Factually, the majority of criteria of predictive ability QSPR/or QSAR is based o the distribution of dots in the plot experimental vs. calculated values of an endpoint. However, the IIC is a new approach which considers the use to both correlation coefficient as well as the distribution of the https://doi.org/10.1016/j.foodres.2019.03.067 Received 1 January 2019; Received in revised form 9 March 2019; Accepted 28 March 2019 ⁎ Corresponding author. E-mail address: pgrachary@soa.ac.in (P.G.R. Achary). Food Research International 122 (2019) 40–46 Available online 29 March 2019 0963-9969/ © 2019 Published by Elsevier Ltd. T