Machine learning‐based atom contribution method for the prediction of surface charge density profiles and solvent design

Abstract

Solvents are widely used in chemical processes. The use of efficient model‐based solvent selection techniques is an option worth considering for rapid identification of candidates with better economic, environment and human health properties. In this paper, an optimization‐based MLAC‐CAMD framework is established for solvent design, where a novel machine learning‐based atom contribution method is developed to predict molecular surface charge density profiles (σ‐profiles). In this method, weighted atom‐centered symmetry functions are associated with atomic σ‐profiles using a high‐dimensional neural network model, successfully leading to a higher prediction accuracy in molecular σ‐profiles and better isomer identifications compared with group contribution methods. The new method is integrated with the computer‐aided molecular design technique by formulating and solving a mixed‐integer nonlinear programming model, where model complexities are managed with a decomposition‐based strategy. Finally, two case studies involving crystallization and reaction are presented to highlight the wide applicability and effectiveness of the MLAC‐CAMD framework.

An accurate and interpretable deep learning model for environmental properties prediction using hybrid molecular representations

“…Environmental benefit, which is one of the three pillars of sustainable development, 1 promotes the application of green chemistry and moves the chemical industry forward 2–6 . Many environmental impact factors (such as toxicity, CO 2 emissions, and water consumption) should be incorporated in molecule design, 7–9 solvent selection 10–12 as well as chemical synthesis 13–16 . As a fundamental environmental property, lipophilicity plays a vital role in governing kinetic and thermodynamics aspects of organic compound reaction 17 and provides crucial information on absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) 18–20 …”

Crystallization solvent design based on a new quantitative prediction model of crystal morphology

“…Since the CAMD method was proposed for designing extractant for the liquid–liquid extraction process, 11 it has been widely used in extractive, 12,13 absorption, 14–16 distillation, 17 crystallization, 7,18 and many other fields 19,20 . Karunanithi et al 4,5 used CAMD method to design suitable crystallization solvents for ibuprofen and other carboxylic acids, taking the maximum crystal yield as the objective function and considering related solvent properties (such as normal melting point, normal boiling point, miscibility) and the solid–liquid equilibrium equation.…”

A systematic modeling methodology of deep neural network‐based structure‐property relationship for rapid and reliable prediction on flashpoints

“…The QSPR modeling on a set of structure‐related chemicals aims to build a mathematical correlation between molecular structures and quantitative chemical attributes. Many QSPR models are built with various mathematical algorithms, such as multiple linear regression, 3–7 artificial neural networks (ANN), 5,8–12 and support vector machine 13–15 . In the fields of biology and chemistry, ANN demonstrates excellent capabilities in nonlinear function approximation, especially when facing multidimensional regression problems.…”

“…In the fields of biology and chemistry, ANN demonstrates excellent capabilities in nonlinear function approximation, especially when facing multidimensional regression problems. To date, ANN is widely used in QSPRs 5,8,11,12,16–18 thanks to its excellent ability in adaptive learning, nonlinear fitting and processing multidimensional inputs. Pan et al 9,10 developed several models to estimate FPT based on the back‐propagation neural networks (BPNN) for 44 alkanes, 40 fatty alcohols and 92 alkanes.…”