A Deep Neural Network for Accurate and Robust Prediction of the Glass Transition Temperature of Polyhydroxyalkanoate Homo- and Copolymers

Abstract

The purpose of this study was to develop a data-driven machine learning model to predict the performance properties of polyhydroxyalkanoates (PHAs), a group of biosourced polyesters featuring excellent performance, to guide future design and synthesis experiments. A deep neural network (DNN) machine learning model was built for predicting the glass transition temperature, Tg, of PHA homo- and copolymers. Molecular fingerprints were used to capture the structural and atomic information of PHA monomers. The other input variables included the molecular weight, the polydispersity index, and the percentage of each monomer in the homo- and copolymers. The results indicate that the DNN model achieves high accuracy in estimation of the glass transition temperature of PHAs. In addition, the symmetry of the DNN model is ensured by incorporating symmetry data in the training process. The DNN model achieved better performance than the support vector machine (SVD), a nonlinear ML model and least absolute shrinkage and selection operator (LASSO), a sparse linear regression model. The relative importance of factors affecting the DNN model prediction were analyzed. Sensitivity of the DNN model, including strategies to deal with missing data, were also investigated. Compared with commonly used machine learning models incorporating quantitative structure–property (QSPR) relationships, it does not require an explicit descriptor selection step but shows a comparable performance. The machine learning model framework can be readily extended to predict other properties.

Control of D-lactic acid content in P(LA-3HB) copolymer in the yeast Saccharomyces cerevisiae using a synthetic gene expression system

“…The increased flexibility and lower melting temperatures are not limited to P(LA-3HB) copolymer, but similar phenomena are observed also with PHA copolymers containg 3-hydroxypropionate and 4-hydroxybutyrate monomers ( Doi et al, 1990 ; Li et al, 2010 ; Meng et al, 2012 ). In fact, a recent PHA modelling study, focusing on predicting the glass transition temperature (Tg) of different PHA copolymers, reports that the relative amount of two different monomers is the second most important parameter defining the Tg, after the choice of the monomer ( Jiang et al, 2020 ). These findings emphasise the importance of controlling the monomer ratios in development of the new PHA copolymers.…”

Bioplastic Design using Multitask Deep Neural Networks

“…3 Also, PHAs with side-chain-terminating phenyl groups exhibit higher T g s because of increased rigidity due to enhanced interchain interactions resulting from the polar side chain functional groups. 15,18 Besides systematic structural and chemical alterations, copolymers provide an additional knob to grow the accessible property space by not only combining multiple PHA-based motifs but also PHAs with conventional polymers. 19 In the past, PHA-PHA copolymers have been found to improve mechanical properties while keeping high T m and low T g values, which is ideal for applications that require large temperature operation windows.…”

“…The burgeoning field of polymer informatics [21][22][23][24][25][26] offers an exciting alternative route to address such search problems by using modern data-driven machine learning approaches. 27,28 The present study, with the details of the workflow and machine learning framework outlined in Figure 1, has several vital elements. First, we develop efficient multitask deep neural network-based multiproperty predictors for copolymers that forecast three different thermal (T g , T m , and T d ), four different mechanical (E, σ y , σ b , and b ), and six gas permeability…”

Bioplastic Design using Multitask Deep Neural Networks

“…The burgeoning field of polymer informatics [21][22][23][24][25][26] offers an exciting alternative route to address such search problems by using modern data-driven machine learning approaches. 27,28 The present study, with the details of the workflow and machine learning framework outlined in Figure 1, has several vital elements. First, we develop efficient multitask deep neural network-based multiproperty predictors for copolymers that forecast three different thermal (T g , T m , and T d ), four different mechanical (E, σ y , σ b , and ǫ b ), and six gas permeability (µ g |g ∈ {O 2 , CO 2 , N 2 , H 2 , He, CH 4 }) properties using nearly 23 000 experimental data points pertaining to a diverse range of homo-and copolymer chemistries.…”