Learning to Make Chemical Predictions: The Interplay of Feature Representation, Data, and Machine Learning Methods

Haghighatlari, Mojtaba; Li, Jie; Heidar‐Zadeh, Farnaz; Liu, Yuchen; Guan, Xingyi; Head‐Gordon, Teresa

doi:10.1016/j.chempr.2020.05.014

Cited by 85 publications

(75 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, long MD simulations generate a large amount of high-dimensional data, making it difficult to identify which region or residues of the RBD make significant contributions to these differences. 33,34,36 ML approaches can be excellent tools for identifying differences in MD trajectories 34,35,37,42 and have recently been applied in computational studies related to SARS-CoV-2. 35,43,44 Inspired by the work of Fleetwood et al, 34 we trained ML classifiers to distinguish between the configurations from the SARS-CoV and SARS-CoV-2 trajectories, and, in the process, rate the importance of each feature to the classification.…”

Section: Identification Of Residues Distinguishing Sars-cov From Sarsmentioning

confidence: 99%

“…These simulations produce an extraordinary amount of data, which poses serious challenges for analysis and interpretation. [33][34][35][36][37][38] Similar to the recent work of Fleetwood et al, 34 we use supervised machine learning (ML) approaches to assist in identifying those residues that contribute the most to dynamical differences between the two viral RBDs. Based on this identification, we further quantify the relative changes in binding free energy for the RBD with ACE2 via free energy perturbation (FEP) calculations.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Critical interactions for SARS-CoV-2 spike protein binding to ACE2 identified by machine learning

Pavlova

Zhang

Acharya

et al. 2021

Preprint

View full text Add to dashboard Cite

Both SARS-CoV and SARS-CoV-2 bind to the human ACE2 receptor. Based on high-resolution structures, the two viruses bind in practically identical conformations, although several residues of the receptor-binding domain (RBD) differ between them. Here we have used molecular dynamics (MD) simulations, machine learning (ML), and free energy perturbation (FEP) calculations to elucidate the differences in RBD binding by the two viruses. Although only subtle differences were observed from the initial MD simulations of the two RBD-ACE2 complexes, ML identified the individual residues with the most distinctive ACE2 interactions, many of which have been highlighted in previous experimental studies. FEP calculations quantified the corresponding differences in binding free energies to ACE2, and examination of MD trajectories provided structural explanations for these differences. Lastly, the energetics of emerging SARS-CoV-2 mutations were studied, showing that the affinity of the RBD for ACE2 is increased by N501Y and E484K mutations but is slightly decreased by K417N.

show abstract

Section: Identification Of Residues Distinguishing Sars-cov From Sarsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Critical interactions for SARS-CoV-2 spike protein binding to ACE2 identified by machine learning

Pavlova

Zhang

Acharya

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The factors that affect the performances of prediction models can be basically grouped into four categories ( Fig. 1a), where the first two pertain to the data and the latter two pertain to the model [4]:…”

Section: • Training and Testing The Modelmentioning

confidence: 99%

Pushing the limits of solubility prediction via quality-oriented data selection

Sorkun

Koelman

2020

Preprint

View full text Add to dashboard Cite

Accurate prediction of the solubility of chemical substances in solvents remains a challenge. The sparsity of high-quality solubility data is recognized as the biggest hurdle in the development of robust data-driven methods for practical use. Nonetheless, the effects of the quality and quantity of data on aqueous solubility predictions have not yet been scrutinized. In this study, the roles of the size and the quality of datasets on the performances of the solubility prediction models are unraveled, and the concepts of actual and observed performances are introduced. In an effort to curtail the gap between actual and observed performances, a quality-oriented data selection method, which evaluates the quality of data and extracts the most accurate part of it through statistical validation, is designed. Applying this method on the largest publicly available solubility database and using a consensus machine learning approach, a top-performing solubility prediction model is achieved.

show abstract

“…Accordingly, the importance of the interpretability for machine learning models in chemistry has been outlined before. 102 One pathway towards inherently explainable artificial intelligence could be the re-emergence of symbolic artificial intelligence. 103 In chemistry, interpretability goes hand-in-hand with the representation of a given problem.…”

Section: New Ideas and Paradigmsmentioning

confidence: 99%

Navigating through the Maze of Homogeneous Catalyst Design with Machine Learning

Gomes

Pollice

Aspuru‐Guzik

2020

Preprint

View full text Add to dashboard Cite

<div><div><div><p>The ability to forge difficult chemical bonds through catalysis has transformed society on all fronts, from feeding our ever-growing populations to increasing our life-expectancies through the synthesis of new drugs. However, developing new chemical reactions and catalytic systems is a tedious task that requires tremendous discovery and optimization efforts. Over the past decade, advances in machine learning have revolutionized a whole new way to approach data- intensive problems, and many of these developments have started to enter chemistry. However, similar progress in the field of homogenous catalysis are only in their infancy. In this article, we want to outline our vision for the future of catalyst design and the role of machine learning to navigate this maze.</p></div></div></div>

show abstract

Learning to Make Chemical Predictions: The Interplay of Feature Representation, Data, and Machine Learning Methods

Cited by 85 publications

References 59 publications

Critical interactions for SARS-CoV-2 spike protein binding to ACE2 identified by machine learning

Critical interactions for SARS-CoV-2 spike protein binding to ACE2 identified by machine learning

Pushing the limits of solubility prediction via quality-oriented data selection

Navigating through the Maze of Homogeneous Catalyst Design with Machine Learning

Contact Info

Product

Resources

About