Computer-aided molecular design in the continuous-molecular targeting framework using group-contribution PC-SAFT

Computer-aided molecular design allows generating novel fluids fulfilling a set of target properties. An integrated design of fluid and process directly employs a process-based objective function. In this work, we solve the integrated process and fluid design problem using the continuous-molecular targeting computer-aided molecular design (CoMT–CAMD) framework. CoMT–CAMD exploits the molecular picture underlying the PC-SAFT equation of state. In the simultaneous optimization of process and fluid, relaxed pure component parameters allow for an efficient optimization. The result is a hypothetical optimal target fluid. In previous work, fluids showing similar performance as the target fluid were obtained from a mapping onto a database. Here, we integrate computer-aided molecular design to realize the actual design of novel fluids. The resulting method for fluid design is based on a group-contribution method for the PC-SAFT parameters (GPC-SAFT) and applied to the design of working fluids for Organic Rankine cycles and solvents for CO₂ capture.     

结合图论和基团贡献法的非均相共沸精馏挟带剂设计

非均相共沸精馏挟带剂的计算机辅助分子设计(CAMD)由分子合成、分子筛选及分子确认3个环节递进构成。在分子合成环节,预选基团,限定合成分子的基团总数及类型,基于图论原理实现由基团到分子的自动合成。在分子筛选环节,依据基础物性筛选指标形成基础分子库,输入待分离物系,采用非均相共沸物形成判据筛选出若干候选分子。在分子确认环节,由非均相共沸温度及组成、挟带剂的汽化热等参数组成模糊综合评判函数,实现分子排序,从而输出一组较优挟带剂。以乙酸-水物系、乙腈-乙酸乙酯物系为例,得到了相应设计结果,与文献结果进行了对比。研究表明该方法及所编程序具备可靠实用性,可为近沸程及共沸混合物分离过程的开发与设计提供先导性支持。     

A systematic methodology for optimal mixture design in an integrated biorefinery

Biomass is a sustainable source of energy which can be utilised to produce value-added products such as biochemical products and biomaterials. In order to produce a sustainable supply of such value-added products, an integrated biorefinery is required. An integrated biorefinery is a processing facility that integrates multiple biomass conversion pathways to produce value-added products. To date, various biomass conversion pathways are available to convert biomass into a wide range of products. Due to the large number of available pathways, various systematic screening tools have been developed to address the process design aspect of an integrated biorefinery. Process design however, is often inter-linked with product design as it is important to identify the optimal molecule (based on desired product properties) prior to designing its optimal production routes. In cases where the desired product properties cannot be met by a single component chemical product, a mixture of chemicals would be required. In this respect, product and process design decisions would be a challenging task for an integrated biorefinery. In this work, a novel two-stage optimisation approach is developed to identify the optimal conversion pathways in an integrated biorefinery to convert biomass into the optimal mixtures in terms of target product properties. In the first stage, the optimal mixture is designed via computer-aided molecular design (CAMD) technique. CAMD technique is a reverse engineering approach which predicts the molecules with optimal properties using property prediction models. Different classes of property models such as group contribution (GC) models and quantitative structure property relationship (QSPR) are adapted in this work. The main component of the mixture is first determined from the target product properties. This is followed by the identifying of additive components to form an optimal mixture with the main component based on the desired product properties. Once the optimal mixture is determined, the second stage identifies the optimal conversion pathways via superstructural mathematical optimisation approach. With such approach, the optimal conversion pathways can be determined based on different optimisation objectives (e.g. highest product yield, lowest environmental impact etc.). To illustrate the proposed methodology, a case study on the design of fuel additives as a mixture of different molecules from palm-based biomass is presented. With the developed methodology, optimal fuel additives are designed based on optimal target properties. Once the optimal fuel additives are designed, the optimal conversion pathways in terms of highest product yield and economic performance that convert biomass into the optimal fuel additives are identified.     

V-Dock: Fast Generation of Novel Drug-like Molecules Using Machine-Learning-Based Docking Score and Molecular Optimization

We propose a computational workflow to design novel drug-like molecules by combining the global optimization of molecular properties and protein-ligand docking with machine learning. However, most existing methods depend heavily on experimental data, and many targets do not have sufficient data to train reliable activity prediction models. To overcome this limitation, protein-ligand docking calculations must be performed using the limited data available. Such docking calculations during molecular generation require considerable computational time, preventing extensive exploration of the chemical space. To address this problem, we trained a machine-learning-based model that predicted the docking energy using SMILES to accelerate the molecular generation process. Docking scores could be accurately predicted using only a SMILES string. We combined this docking score prediction model with the global molecular property optimization approach, MolFinder, to find novel molecules exhibiting the desired properties with high values of predicted docking scores. We named this design approach V-dock. Using V-dock, we efficiently generated many novel molecules with high docking scores for a target protein, a similarity to the reference molecule, and desirable drug-like and bespoke properties, such as QED. The predicted docking scores of the generated molecules were verified by correlating them with the actual docking scores.     

Challenges and opportunities in computer-aided molecular design

In this paper, the significant development, current challenges and future opportunities in the field of chemical product design using computer-aided molecular design (CAMD) tools are highlighted. With the gaining of focus on the design of novel and improved chemical products, the traditional heuristic based approaches may not be effective in designing optimal products. This leads to the vast development and application of CAMD tools, which are methods that combine property prediction models with computer-assisted search in the design of various chemical products. The introduction and development of different classes of property prediction methods in the overall product design process is discussed. The exploration and application of CAMD tools in numerous single component product designs, mixture design, and later in the integrated process-product design are reviewed in this paper. Difficulties and possible future extension of CAMD are then discussed in detail. The highlighted challenges and opportunities are mainly about the needs for exploration and development of property models, suitable design scale and computational effort as well as sustainable chemical product design framework. In order to produce a chemical product in a sustainable way, the role of each level in a chemical product design enterprise hierarchy is discussed. In addition to process parameters and product quality, environment, health and safety performance are required to be considered in shaping a sustainable chemical product design framework. On top of these, recent developments and opportunities in the design of ionic liquids using molecular design techniques have been discussed.     

Molecular Conditional Generation and Property Analysis of Non-Fullerene Acceptors with Deep Learning

The proposition of non-fullerene acceptors (NFAs) in organic solar cells has made great progress in the raise of power conversion efficiency, and it also broadens the ways for searching and designing new acceptor molecules. In this work, the design of novel NFAs with required properties is performed with the conditional generative model constructed from a convolutional neural network (CNN). The temporal CNN is firstly trained to be a good string-based molecular conditional generative model to directly generate the desired molecules. The reliability of generated molecular properties is then demonstrated by a graph-based prediction model and evaluated with quantum chemical calculations. Specifically, the global attention mechanism is incorporated in the prediction model to pool the extracted information of molecular structures and provide interpretability. By combining the generative and prediction models, thousands of NFAs with required frontier molecular orbital energies are generated. The generated new molecules essentially explore the chemical space and enrich the database of transformation rules for molecular design. The conditional generation model can also be trained to generate the molecules from molecular fragments, and the contribution of molecular fragments to the properties is subsequently predicted by the prediction model.     

Enhanced Deep-Learning Prediction of Molecular Properties via Augmentation of Bond Topology

Deep learning has made great strides in tackling chemical problems, but still lacks full-fledged representations for three-dimensional (3D) molecular structures for its inner working. For example, the molecular graph, commonly used in chemistry and recently adapted to the graph convolutional network (GCN), is inherently a 2D representation of 3D molecules. Herein we propose an advanced version of the GCN, called 3DGCN, which receives 3D molecular information from a molecular graph augmented by information on bond direction. While outperforming state-of-the-art deep-learning models in the prediction of chemical and biological properties, 3DGCN has the ability to both generalize and distinguish molecular rotations in 3D, beyond 2D, which has great impact on drug discovery and development, not to mention the design of chemical reactions.     

Optimal design of solvents for extractive reaction processes

It is well known that solvents can have significant effects on rates and equilibrium compositions of chemical reactions. The computer‐aided molecular design (CAMD) of solvents for heterogeneous liquid phase reactions is challenging due to multiple solvent effects on reaction and phase equilibria. In this work, we propose a CAMD methodology based on a genetic algorithm (GA) for identifying optimal solvents for liquid phase reactions where the objective is to maximize the reaction equilibrium conversion. In particular, a novel molecular encoding method is introduced to facilitate the construction and evaluation of solvent molecules in a defined structure space. The reliability of the method for fast identification of optimal reaction solvents is demonstrated for a selected biphasic esterification reaction. The proposed approach opens up new perspectives for intensifying extractive reaction processes via the purposeful design of solvent molecules. © 2016 American Institute of Chemical Engineers AIChE J, 62: 3238–3249, 2016     

基团贡献法分子设计研究的进展

利用基团贡献法可预测化合物的性质,还能用于化合物的计算机辅助分子设计(CAMD).本文论述了基于基团贡献法CAMD的基本原理,以及在溶剂和聚合物等领域分子设计的应用,对分子设计的计算方法也作了简单的介绍.随着绿色溶剂和新型聚合物材料需求的增加,基团贡献法CAMD将大有应用前景.     

COSMO‐based computer‐aided molecular/mixture design: A focus on reaction solvents

In this article, we investigate reaction solvent design using COSMO‐RS thermodynamics in conjunction with computer‐aided molecular design (CAMD) techniques. CAMD using COSMO‐RS has the distinct advantage of being a method based in quantum chemistry, which allows for the incorporation of quantum‐level information about transition states, reactive intermediates, and other important species directly into CAMD problems. This work encompasses three main additions to our previous framework for solvent design (Austin et al., Chem Eng Sci. 2017;159:93–105): (1) altering the group contribution method to estimate hydrogen‐bonding and non‐hydrogen‐bonding σ‐profiles; (2) ab initio modeling of strong solute/solvent interactions such as H‐bonding or coordinate bonding; and (3) solving mixture design problems limited to common laboratory and industrial solvents. We apply this methodology to three diverse case studies: accelerating the reaction rate of a Menschutkin reaction, controlling the chemoselectivity of a lithiation reaction, and controlling the chemoselectivity of a nucleophilic aromatic substitution reaction. We report improved solvents/mixtures in all cases. © 2017 American Institute of Chemical Engineers AIChE J, 63: 104–122, 2018     

Generic mathematical programming formulation and solution for computer-aided molecular design

This short communication presents a generic mathematical programming formulation for computer-aided molecular design (CAMD). A given CAMD problem, based on target properties, is formulated as a mixed integer linear/non-linear program (MILP/MINLP). The mathematical programming model presented here, which is formulated as an MILP/MINLP problem, considers first-order and second-order molecular groups for molecular structure representation and property estimation. It is shown that various CAMD problems can be formulated and solved through this model.     

A computer-aided molecular design framework for crystallization solvent design

One of the key decisions in designing solution crystallization processes is the selection of solvents. In this paper, we present a computer-aided molecular design (CAMD) framework for the design and selection of solvents and/or anti-solvents for solution crystallization. The CAMD problem is formulated as a mixed integer nonlinear programming (MINLP) model. Although, the model allows any combination of performance objectives and property constraints, in the case studies, potential recovery was considered as the performance objective. The latter, needs to be maximized, while other solvent property requirements such as solubility, crystal morphology, flashpoint, toxicity, viscosity, normal boiling and melting point are posed as constraints. All the properties are estimated using group contribution methods. The MINLP model is then solved using a decomposition approach to obtain optimal solvent molecules. Solvent design and selection for two types of solution crystallization processes namely cooling crystallization and drowning out crystallization are presented. In the first case study, the design of single compound solvent for crystallization of ibuprofen, which is an important pharmaceutical compound, is addressed. One of the important issues namely, the effect of solvent on the shape of ibuprofen crystals is also considered in the MINLP model. The second case study is a mixture design problem where an optimal solvent/anti-solvent mixture is designed for crystallization of ibuprofen by the drowning out technique. For both case studies the performance of the solvents are verified qualitatively through SLE diagrams.     

Efficient ant colony optimization for computer aided molecular design: Case study solvent selection problem

In this paper, we propose a novel computer-aided molecular design (CAMD) methodology for the design of optimal solvents based on an efficient ant colony optimization (EACO) algorithm. The molecular design problem is formulated as a mixed integer nonlinear programming (MINLP) model in which a solvent performance measure is maximized (solute distribution coefficient) subject to structural feasibility, property, and process constraints. In developing the EACO algorithm, the better uniformity property of Hammersley sequence sampling (HSS) is exploited. The capabilities of the proposed methodology are illustrated using a real world case study for the design of an optimal solvent for extraction of acetic acid from waste process stream using liquid–liquid extraction. The UNIFAC model based on the infinite dilution activity coefficient is used to estimate the mixture properties. New solvents with better targeted properties are proposed.     

An experimental verification of morphology of ibuprofen crystals from CAMD designed solvent

In our previous work [Karunanithi et al., 2006. A computer-aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61, 1247-1260] we proposed a computer-aided molecular design (CAMD) framework to design solvents for crystallization processes. One of the important aspects of that work was the consideration of a qualitative property, namely crystal morphology, along with other physico-chemical properties (quantitative) of the solvents within the modeling framework. However, it is our view that consideration of any qualitative property, such as morphology of crystals formed from solvents, necessitates additional experimental verification steps. In this work we report the experimental verification of crystal morphology for the case study, solvent design for ibuprofen crystallization, presented in Karunanithi et al. [2006. A computer-aided molecular design framework for crystallization solvent design. Chemical Engineering Science 61, 1247-1260]. This we believe is an important step for the validation of the proposed solvent design model.     

Iterative Process Design with Surrogate-Assisted Global Flowsheet Optimization

Flowsheet optimization is an important part of process design where commercial process simulators are widely used, due to their extensive library of models and ease of use. However, the application of a framework for global flowsheet optimization upon them is computationally expensive. Based on machine learning methods, we added mechanisms for rejection and generation of candidates to a framework for global flowsheet optimization. These extensions halve the amount of time needed for optimization such that the integration of the framework in a workflow for iterative process design becomes applicable.     

基于分子动力学的橡胶聚合物计算机辅助设计方法

聚合物分子设计的关键步骤是得到能够满足多种性质要求的重复单元结构。作为化学产品工程中的新型发展手段，计算机辅助分子设计(CAMD)技术可以通过基团贡献法生成满足约束条件的聚合物重复单元结构，分子动力学(MD)技术则可以在微观层面上进行计算机实验模拟系统性质。建立了聚合物的CAMD-MD通用设计方法，并进行轮胎橡胶聚合物的分子设计，首先基于基团贡献法进行重复单元的设计；其次，利用层次分析法确定多性质权重排名，并基于分子动力学方法探究候选结构的性质；最后将方法应用于实际橡胶结构中，模拟得到聚能密度、密度、玻璃化转换温度和热导率性质，验证了方法的可行性。     

Designing optical glasses by machine learning coupled with a genetic algorithm

Engineering new glass compositions have experienced a sturdy tendency to move forward from (educated) trial-and-error to data- and simulation-driven strategies. In this work, we developed a computer program that combines data-driven predictive models (in this case, neural networks) with a genetic algorithm to design glass compositions with desired combinations of properties. First, we induced predictive models for the glass transition temperature (T_g) using a dataset of 45,302 compositions with 39 different chemical elements, and for the refractive index (n_d) using a dataset of 41,225 compositions with 38 different chemical elements. Then, we searched for relevant glass compositions using a genetic algorithm informed by a design trend of glasses having high n_d (1.7 or more) and low T_g (500 °C or less). Two candidate compositions suggested by the combined algorithms were selected and produced in the laboratory. These compositions are significantly different from those in the datasets used to induce the predictive models, showing that the used method is indeed capable of exploration. Both glasses met the constraints of the work, which supports the proposed framework. Therefore, this new tool can be immediately used for accelerating the design of new glasses. These results are a stepping stone in the pathway of machine learning-guided design of novel glasses.     

基于高阶基团贡献法与COSMO-SAC模型的溶剂设计方法

提出了一种基于高阶基团贡献法与类导体屏蔽片段活度系数模型（conductor like screening model-segment activity coefficient, COSMO-SAC）的计算机辅助溶剂设计方法（computer-aided molecular design, CAMD）。首先,基于高阶基团贡献法（higher-order group contribution, GC⁺）与COSMO-SAC模型构建GC⁺-COSMO方法,关联分子基团组合与表面屏蔽电荷密度分布[σ-profiles, p(σ)]、分子空腔体积V_c,实现对二者的高通量预测;然后结合基于简化分子线性输入系统（simplified molecular input line entry system, SMILES）的异构体生成算法与GC⁺-COSMO方法实现CAMD技术对异构体的识别及性质区分;最后,通过目标函数与约束方程组成的混合整数非线性规划模型（mixed integer nonlinear programming, MINLP）来建立溶剂设计问题,进一步采用分解式算法优化求解,实现溶剂优化设计目标。基于以上模型和方法开展了狄尔斯-阿尔德（Diels-Alder, DA）竞争性反应溶剂设计,验证了提出的方法的可行性与有效性。