Publications

A Review of Molecular Representation in the Age of Machine Learning

WIREs Computational Molecular Science, 2022

Research in chemistry increasingly requires interdisciplinary work prompted by, among other things, advances in computing, machine learning, and artificial intelligence. Everyone working with molecules, whether chemist or not, needs an understanding of the representation of molecules in a machine-readable format, as this is central to computational chemistry. Four classes of representations are introduced: string-, connection table-, feature based-, and computer learned-representations. Three of the most significant representations are SMILES, InChI, and the MDL molfile, of which SMILES was the first to successfully be used in conjunction with a variational autoencoder to yield a continuous representation of molecules. This is noteworthy because a continuous representation allows for efficient navigation of the immensely large chemical space of possible molecules. Since 2018, when the first model of this type was published, considerable effort has been put into developing novel and improved methodologies. Most, if not all, researchers in the community make their work easily accessible on GitHub, though discussion of computation time and domain of applicability is often overlooked. Herein we present questions for consideration in future work which we believe will make chemical variational autoencoders even more accessible.

Multi-task Bayesian Optimisation of Chemical Reactions

NeurIPS Machine Learning for Molecules, 2020

Recent work has shown how Bayesian optimization (BO) is an efficient method for optimizing expensive experiments such as chemical reactions. However, in previous studies, each optimization has been started from scratch with no information about previous or similar chemical optimization studies. Therefore, BO can still require more iterations than many experimental budgets provide. Here, we overcome this challenge using multi-task BO. Through in silico benchmarking studies, we show how past experimental data can be leveraged to improve the quality and speed of reaction optimization.

A Framework for Biogas Exploitation in Italian Waste Water Treatment Plants

Chemical Engineering Transactions, 2019

Effective utilisation of biogas is an important step in increasing usage of renewable energy, due to the great flexibility that solar and wind power in particular lacks. Biogas generated through anaerobic digestion (AD) of sewage sludge addresses environmental concerns together with creating electricity generation potential. There is currently no optimisation-based decision-support framework to determine the best use of biogas from a Waste Water Treatment Plant (WWTP), and provide a market outlook for each of the options. This work proposes a novel multi-period Mixed Integer Linear Program (MILP) model for dispatch and selection of technologies capable of exploiting biogas produced from sludge. The novelty is also highlighted by extrapolating the optimised results to a broader analysis of 855 Italian WWTPs with Population Equivalent (P.E.) > 20,000. The use of real input data provides a unique added value to the work. The modelling framework is applied to several case studies. Results show that 7–23 % savings in operating costs are possible from integrating three systems to exploit biogas, and the trade-offs between capital and operating costs affect the optimal system choice. Furthermore, market driven scenarios are used to analyse how to improve the economic performance.