Hybrid, Interpretable Machine Learning for Thermodynamic Property Estimation using Grammar2vec for Molecular Representation

Mann, Vipul; Brito, Karoline Dantas; Gani, Rafiqul; Venkatasubramanian, Venkat

doi:10.1016/j.fluid.2022.113531

Cited by 17 publications

(11 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A detailed description of the SVM model and the mathematical framework that underlies it is provided in ref. 2.…”

Section: Application In Reaction Class Prediction Problemmentioning

confidence: 99%

“…Hybrid AI models have a central role to play in driving chemistry growth by combining domain knowledge in the form of symbolic AI with numeric machine learning methods, 1 thus leveraging the expertise of a chemist and the numeric stronghold of AI methods. Consequently, several hybrid AI-based methods have been reported for problems including thermodynamic property estimation, 2,3 reaction prediction and retrosynthesis, 4,5 and chemical product design among several others as presented in the excellent review articles. 6–9…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

AI-driven hypergraph network of organic chemistry: network statistics and applications in reaction classification

Mann

Venkatasubramanian

2023

React. Chem. Eng.

Self Cite

View full text Add to dashboard Cite

show abstract

“…A detailed description of the SVM model and the mathematical framework that underlies it is provided in ref. 2.…”

Section: Application In Reaction Class Prediction Problemmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

AI-driven hypergraph network of organic chemistry: network statistics and applications in reaction classification

Mann

Venkatasubramanian

2023

React. Chem. Eng.

Self Cite

View full text Add to dashboard Cite

show abstract

“…GNN has also been shown to be effective to capture the underlying thermodynamics for the prediction of infinite dilution activity coefficients [193]. Some of the recent notable contributions regarding simplified molecular-input line-entry system (SMILES)-based molecular representations include the work by Hirohara et al [194], Krenn et al [195], and the framework for molecular representation which was used to develop interpretable ML to predict thermodynamic properties [196]. Deep neural networks (DNNs) are particularly effective at correlating high dimensional complex relationships [197].…”

Section: Challenges and Future Research Directionsmentioning

confidence: 99%

“…194, Krenn et al. 195, and the framework for molecular representation which was used to develop interpretable ML to predict thermodynamic properties 196. Deep neural networks (DNNs) are particularly effective at correlating high dimensional complex relationships 197.…”

Section: Challenges and Future Research Directionsmentioning

confidence: 99%

An Overview of Computer‐aided Molecular and Process Design

Iftakher

Monjur

Hasan

2023

Chemie Ingenieur Technik

View full text Add to dashboard Cite

Identifying sustainable chemical processes often depends on the choice of enabling materials that directly influence the overall performance. Matching property targets while incorporating adequate process knowledge is essential for optimal material selection. Multi-scale decisions need to be taken simultaneously to determine the optimal process configurations, operating conditions, and material structures. Integrating molecular to process scale decisions within an equation-oriented optimization framework leads to large-scale mixed-integer nonlinear programs (MINLP). Over the years, several solution approaches have been suggested to tackle this issue. Here, the current state-of-the-art in the field of computer-aided molecular and process design (CAMPD) is discussed and key challenges and open questions are highlighted that may stimulate future research.

show abstract

“…One of them uses SMILES, 21 the de facto standard for cheminformatics I/O, to convert molecules into one-dimensional character sequences. In this way, QSPR modeling is converted into a natural language processing (NLP) problem that can be solved using NLP techniques such as recurrent neural networks (RNNs) 22,23 or transformers. 24,25 Similar methods have achieved remarkable performance, even in synthesis planning problems, 26,27 which are considered more complex and difficult.…”

Section: Introductionmentioning

confidence: 99%

Improved graph‐based multitask learning model with sparse sharing for quantitative structure–property relationship prediction of drug molecules

Bai

Yuan

2022

AIChE Journal

View full text Add to dashboard Cite

The quantitative structure–property relationship (QSPR) is a fundamental technique for evaluating and screening potentially valuable molecules in the field of drug discovery. There is an urgent need to speed up pharmaceutical research and development and a huge chemical space to explore, which necessitate effective and precise computer‐aided QSPR modeling methods. Previous studies with various deep learning models are limited because they are trained on separate small datasets, known as the small‐sample problem. Using transfer learning, this article describes a sparse sharing method that uses advanced graph‐based models to construct an efficient and reasonable multitask learning workflow for QSPR prediction. The proposed workflow is systematically and comprehensively tested with four benchmark datasets containing different targets, and several precisely predicted molecular examples are illustrated. The results demonstrate that an obvious improvement in the prediction of molecular properties is achieved, along with the ability to predict multiple properties simultaneously.

show abstract

Hybrid, Interpretable Machine Learning for Thermodynamic Property Estimation using Grammar2vec for Molecular Representation

Cited by 17 publications

References 39 publications

AI-driven hypergraph network of organic chemistry: network statistics and applications in reaction classification

AI-driven hypergraph network of organic chemistry: network statistics and applications in reaction classification

An Overview of Computer‐aided Molecular and Process Design

Improved graph‐based multitask learning model with sparse sharing for quantitative structure–property relationship prediction of drug molecules

Contact Info

Product

Resources

About