Designing deep eutectic solvents for efficient CO2 capture: A data-driven screening approach

Makarov, Dmitriy M.; Fadeeva, Yu. A.; Golubev, Vasiliy A.; Kolker, A. M.

doi:10.1016/j.seppur.2023.124614

Cited by 14 publications

(3 citation statements)

References 80 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This ensures that each subset is utilized four times for training and once for validation. In ML modeling, when dealing with molecular compounds that exhibit changes in properties under different temperature and/or pressure conditions, two distinct cross-validation strategies can be employed. , Similar to training/test data splitting, the same two approaches were employed for model validation, namely, “data-points” splitting and “drug” splitting (Figure ). The key distinction between these approaches is that in the “drug” splitting protocol, the folds exclusively contain new drug components structures, whereas in the “data-points” splitting, the same drug structures with different solubilities under various state parameters can be assigned to different folds.…”

Section: Methodsmentioning

confidence: 99%

Prediction of Drug-like Compounds Solubility in Supercritical Carbon Dioxide: A Comparative Study between Classical Density Functional Theory and Machine Learning Approaches

Makarov,

Kalikin,

Budkov

2024

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Supercritical carbon dioxide (scCO 2 ) plays an essential role in various technological procedures, making the solubility of drugs in scCO 2 a crucial aspect of the drug formulation process. This study focuses on utilizing theoretical approaches to predict the solubility of drug-like compounds in scCO2 in order to select the optimum parameters for subsequent experimental procedures. Several machine learning models were developed and compared with a previously established theoretical approach based on classical density functional theory (cDFT). The CatBoost model, utilizing alvaDesc descriptors, demonstrated reasonably accurate predictions for the solubility of 187 drugs (AARD = 1.8%). Meanwhile, the CatBoost model, incorporating CDK descriptors and melting points of drugs as input parameters, exhibited satisfactory accuracy (AARD = 14.3%) in extrapolating predictions for new compounds. Comparing the results between the machine learning approach and the cDFT-based one revealed, on average, a higher accuracy and faster prediction speed for the former. However, cDFT demonstrated a more physical behavior of solubility isotherms compared with the machine learning models. This was particularly evident when the ML models struggled to accurately extrapolate solubility values beyond the experimental range of parameters in the supercritical state. Model CatBoost/CDK is freely accessible at http://chem-predictor.isc-ras.ru/ individual/scco/.

show abstract

Section: Methodsmentioning

confidence: 99%

Prediction of Drug-like Compounds Solubility in Supercritical Carbon Dioxide: A Comparative Study between Classical Density Functional Theory and Machine Learning Approaches

Makarov,

Kalikin,

Budkov

2024

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Machine learning (ML) techniques show promise for accurately and effectively predicting the properties of chemical compounds. ,− In different fields, ML methods , are as accurate as traditional simulation techniques such as MD but require less computing power. , Molecular descriptor–based ML (DBML) models have been used to forecast the MPs of ILs. , Molecular descriptors are the numerical values that feature the molecular structures of the ILs. One of the earlier descriptors used group contribution, , which split molecules into fragments; the MP of an IL is calculated by summing the contribution of each fragment.…”

Section: Introductionmentioning

confidence: 99%

Prediction and Interpretability of Melting Points of Ionic Liquids Using Graph Neural Networks

Feng,

Qin,

Zhang

et al. 2024

ACS Omega

View full text Add to dashboard Cite

Ionic liquids (ILs) have wide and promising applications in fields such as chemical engineering, energy, and the environment. However, the melting points (MPs) of ILs are one of the most crucial properties affecting their applications. The MPs of ILs are affected by various factors, and tuning these in a laboratory is time-consuming and costly. Therefore, an accurate and efficient method is required to predict the desired MPs in the design of novel targeted ILs. In this study, three descriptor-based machine learning (DBML) models and eight graph neural network (GNN) models were proposed to predict the MPs of ILs. Fingerprints and molecular graphs were used to represent molecules for the DBML and GNNs, respectively. The GNN models demonstrated performance superior to that of the DBML models. Among all of the examined models, the graph convolutional model exhibited the best performance with high accuracy (root-mean-squared error = 37.06, mean absolute error = 28.79, and correlation coefficient = 0.76). Benefiting from molecular graph representation, we built a GNN-based interpretable model to reveal the atomistic contribution to the MPs of ILs using a data-driven procedure. According to our interpretable model, amino groups, S + , N + , and P + would increase the MPs of ILs, while the negatively charged halogen atoms, S − , and N − would decrease the MPs of ILs. The results of this study provide new insight into the rapid screening and synthesis of targeted ILs with appropriate MPs.

show abstract

“…Perhaps one of the biggest limitations of these methods is that many machine-learning models provide limited generalization capacity, potentially leading to inaccurate predictions when presented new chemical functionalities . In this regard, it is important to highlight a few DES-related databases recently published by Omar et al, Shi et al, Lavrinenko et al, Hou et al, Hopkins et al, Abdollahzadeh et al, and Makarov et al that can greatly facilitate the advancement of these methods. It is important to note that among other limitations, these databases typically include only a small number of nontraditional DES (such as those based on halogen bonding, X-DES), severely impairing the accuracy of any predictions related to these systems.…”

Section: Introductionmentioning

confidence: 99%

eutXG: A Machine-Learning Model to Understand and Predict the Melting Point of Novel X-Bonded Deep Eutectic Solvents

Ayres,

Bandara,

McMillen

et al. 2024

ACS Sustainable Chem. Eng.

View full text Add to dashboard Cite

We present the application of an extreme gradient boosting model (eutXG) to predict the melting point (MP) of deep eutectic solvents (DES). The model is based on XGBoost, a decision tree ensemble based on gradient boosting designed to be highly scalable that enables superior training speed and prediction accuracy. The selected model�trained with molecular fingerprints, molar ratios, and selected chemical descriptors�enabled the prediction of the MPs of DES with an average accuracy of 97.6%, which represents a difference of just ±2.4% with respect to the values reported in the literature. Using SHapley Additive exPlanations (SHAP), further insights into the relative importance of different inputs used to train the machine learning model were identified. Moreover, the generalization ability of the eutXG model was critically assessed by comparing the predicted vs the experimentally determined MP of a series of novel DES based on halogen bonding, developed by mixing tetraalkylammonium triiodide salts (NPe 4 I 3 or NHex 4 I 3 ) with organoiodines, such as 1,2diiodotetrafluorobenzene (o-F 4 DIB), 1,3-diiodotetrafluorobenzene (m-F 4 DIB), or 2,5-diiodothiophene (2,5-DIT), demonstrating its ability to predict the actual melting with a difference of only 2 K. Our results not only reinforce the importance of having (at least some) representative data for the training step to increase the accuracy of the model's predictions but also demonstrate the ability of eutXG to accelerate the development of novel applications for this entirely new class of hydrophobic DES, potentially impacting a wide range of fields from pharmaceuticals to agrochemicals.

show abstract

Designing deep eutectic solvents for efficient CO2 capture: A data-driven screening approach

Cited by 14 publications

References 80 publications

Prediction of Drug-like Compounds Solubility in Supercritical Carbon Dioxide: A Comparative Study between Classical Density Functional Theory and Machine Learning Approaches

Prediction of Drug-like Compounds Solubility in Supercritical Carbon Dioxide: A Comparative Study between Classical Density Functional Theory and Machine Learning Approaches

Prediction and Interpretability of Melting Points of Ionic Liquids Using Graph Neural Networks

eutXG: A Machine-Learning Model to Understand and Predict the Melting Point of Novel X-Bonded Deep Eutectic Solvents

Contact Info

Product

Resources

About