Prediction of the glass transition temperature of (meth)acrylic polymers containing phenyl groups by recursive neural network

Department of Chemistry and Industrial Chemistry, University of Pisa, via Risorgimento 35, 56126 Pisa, Italy
Polymer (Impact Factor: 3.56). 11/2007; 48(24):7121-7129. DOI: 10.1016/j.polymer.2007.09.043


A recursive neural network QSPR model that can take directly molecular structures as input was applied to the prediction of the glass transition temperature of 277 poly(meth)acrylates. This model satisfactorily predicted the chemical–physical properties of high and low molecular weight acyclic compounds. However, side-chain benzene rings are present in about one half of the selected polymers. In order to render cyclic structures, the molecular representation through hierarchical structures was extended by two methods, named group and cycle breaking, respectively. The latter approach exploits standard unique molecular description systems, i.e. Unique SMILES and InChI. In all cases the prediction was very good, with 15–16 K mean absolute error and 19–21 K standard deviation. This result confirms the robustness of our method with respect to the inclusion of different structures. Moreover, the good performance of the cycle breaking representation paves the way for the investigation of data sets that contain a variety of poorly sampled cyclic structures.

Download full-text


Available from: Carlo Bertinetto, Apr 07, 2015
  • Source
    • "The Dow chemical company has been very active in this area and have combined QSPR in the SYNTHIA program with Accelrys software to make a capability for high throughput reverse engineering design of new polymer chemical structures [17]–[19]. Other approaches have involved the use of neural networks [20], [21] and the EVM Method [22]. The EVM method [22] uses energy, volume and mass to predict the Tg values of acrylate and methacrylate polymers with an error of no more than 10%. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The Molecular Operating Environment software (MOE) is used to construct a series of benzoxazine monomers for which a variety of parameters relating to the structures (e.g. water accessible surface area, negative van der Waals surface area, hydrophobic volume and the sum of atomic polarizabilities, etc.) are obtained and quantitative structure property relationships (QSPR) models are formulated. Three QSPR models (formulated using up to 5 descriptors) are first used to make predictions for the initiator data set (n = 9) and compared to published thermal data; in all of the QSPR models there is a high level of agreement between the actual data and the predicted data (within 0.63-1.86 K of the entire dataset). The water accessible surface area is found to be the most important descriptor in the prediction of T(g). Molecular modelling simulations of the benzoxazine polymer (minus initiator) carried out at the same time using the Materials Studio software suite provide an independent prediction of T(g). Predicted T(g) values from molecular modelling fall in the middle of the range of the experimentally determined T(g) values, indicating that the structure of the network is influenced by the nature of the initiator used. Hence both techniques can provide predictions of glass transition temperatures and provide complementary data for polymer design.
    PLoS ONE 01/2013; 8(1):e53367. DOI:10.1371/journal.pone.0053367 · 3.23 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: A recursive neural network (RNN) was used to predict the melting points of several pyridinium-based ionic liquids (ILs). The RNN is a neural network model for processing structured data that allows for the direct handling of chemical compounds as labelled rooted ordered trees. It constitutes a direct approach to quantitative structure–property relationship (QSPR) of ILs, which avoids the use of dedicated molecular descriptors. The adopted representation of molecular structures captures significant topological aspects and chemical functionalities for each molecule in a general and flexible way. Particular emphasis was given to the representation of cyclic moieties. The model was applied to a set of 126 pyridinium bromides; it was validated by splitting the dataset into a disjoint training set (100 compounds) and test set (26 compounds). Comparison with the results obtained by other QSPR approaches on the same dataset is also presented.
    Green Chemistry 03/2008; 10(3). DOI:10.1039/b708123e · 8.02 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Linear and nonlinear quantitative structure–property relationship (QSPR) models for the prediction of refractive indices of polymers were developed based on a diverse dataset with 120 polymers by using multilinear regression analysis and feed-forward artificial neural networks. Descriptors of the polymers were calculated from their corresponding cyclic dimer structures. The dataset was divided into a training set of 100 polymers and a test set of 20 polymers. As the prediction results of the linear model obtained with the training set, the squared correlation coefficient (R2) and the standard error of estimation (s) for the training set are 0.9431 and 0.0157, respectively; the mean relative error (MRE) for the whole dataset using this linear model is 0.79%. Better results were obtained from the nonlinear model: the MREs for the training and test sets are 0.61% (R2 = 0.9713 and s = 0.0118) and 0.76% (R2 = 0.9613 and s = 0.0144 which are calculated from the model trained with the training set), respectively.
    Chemometrics and Intelligent Laboratory Systems 07/2008; 92(2):152-156. DOI:10.1016/j.chemolab.2008.02.006 · 2.32 Impact Factor
Show more