Multiscale approximations of integral equation-based solvation models
No Thumbnail Available
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Alternative Title(s)
Abstract
Die Anwendbarkeit und Kosten quantenchemischer Solvatationsmodelle, wie dem "embedded cluster reference interaction site model" (EC-RISM) werden vor allem durch die Berechnungskosten der Elektronenstruktur, sowie des elektrostatischen Potentials bestimmt, bedingt durch die iterative Lösung dieser Verfahren und der damit einhergehenden Wiederholung der teuren Elektronenstrukturrechnung.
Hierdurch wird ihre Anwendung auf chemische Systeme geringer Größe beschränkt. In dieser Arbeit wird ein neues Solvatationsmodell, ONIOM-EC-RISM, präsentiert mit dem dieser Rechenaufwand für das bisherige EC-RISM Verfahren, durch die Einführung einer multiskalen Approximation der Elektronenstruktur, drastisch reduziert werden kann. Hierdurch erschließt sich die Möglichkeit das EC-RISM-Solvatationsmodell auch auf größere chemische Systeme, wie z.B. Proteinsysteme, anzuwenden.
Neben den für das Verständnis des ONIOM-EC-RISM-Modells relevanten theoretischen Grundlagen additiver und subtraktiver Multiskalenapproximationen, des statistisch-thermodynamischen RISM-Solvatationsmodells, sowie des methodisch verwandten ONIOM-PCM-Solvatationsmodells, werden die Theorie und technische Implementation des neuen Modells umfangreich dargestellt. Darüber hinaus wird aufgezeigt wie zuvor für das EC-RISM-Referenzmodell verwendete empirische Korrekturen in den ONIOM-EC-RISM-Kontext übertragen werden können. Hierbei wird das Größenextrapolationslimit der ONIOM-Methode ausgenutzt, wodurch Korrekturen erhalten werden, die frei von jeglichen Partitionierungsfehlern sind.
Die resultierenden Modelle sind in der Lage, die $\pKa$-Vorhersagequalität des EC-RISM-Referenzmodells für den pKa-Datensatz der SAMPL6-Challenge, einem "blind prediction" Wettbewerb zur Vorhersage thermodynamischer Größen, zu reproduzieren und teilweise zu übertreffen, während gleichzeitig die Gesamtkosten des EC-RISM-Verfahrens drastisch reduziert werden können.
Neben der Validierung des ONIOM-EC-RISM-Verfahrens anhand von pKa-Werten wird demonstriert, wie das ONIOM-EC-RISM-Modell zusätzlich zur Vorhersage chemischer Verschiebungen eines Pentapeptidsystem verwendet werden kann. In diesem Zusammenhang wird ein neuartiger Ansatz vorgestellt der es erlaubt, pH-abhängige chemische Verschiebungen direkt aus spektroskopischen sowie pKa-Vorhersagen zu berechnen. Dies eröffnet erstmalig die Möglichkeit der direkten Modellierung von NMR-Titrationsexperimenten auf Grundlage des EC-RISM-Solvatationsmodells.
The applicability and costs of quantum chemical solvation models, such as the "embedded cluster reference interaction site model" (EC-RISM), are mainly determined by the calculation costs of the electronic structure and the electrostatic potential, due to the iterative solution of these methods and the associated repetition of the expensive electronic structure calculation. This limits their application to chemical systems of small size. In this work, a new solvation model, ONIOM-EC-RISM, is presented which drastically reduces the computational cost of the previous EC-RISM method by introducing a multiscale approximation of the electronic structure. This opens up the possibility of applying the EC-RISM solvation model to larger chemical systems, such as protein systems. In addition to the relevant theoretical foundations of additive and subtractive multiscale approximations, required for the understanding of the ONIOM-EC-RISM model, the statistical-thermodynamical RISM solvation model and the methodologically related ONIOM-PCM solvation model, as well as the theory and technical implementation of the new model are presented in detail. In addition, it is shown how empirical corrections previously used for the EC-RISM reference model can be transferred to the ONIOM-EC-RISM context. Here, the size extrapolation limit of the ONIOM method is utilised, whereby corrections are obtained that are free of any partitioning errors. The resulting models are able to reproduce and partially exceed the pKa prediction quality of the EC-RISM reference model for the pKa dataset of the SAMPL6 challenge, a blind prediction challenge for thermodynamic quantities, while at the same time drastically reducing the overall cost of the EC-RISM method. In addition to the validation of the ONIOM-EC-RISM method on the basis of pKa values, it will be demonstrated how the ONIOM-EC-RISM model can additionally be used to predict chemical shifts of a pentapeptide system. In this context, a novel approach is presented that allows pH-dependent chemical shifts to be calculated directly from spectroscopic and pKa predictions. This opens up for the first time the possibility to directly model NMR titration experiments on the basis of the EC-RISM solvation model.
The applicability and costs of quantum chemical solvation models, such as the "embedded cluster reference interaction site model" (EC-RISM), are mainly determined by the calculation costs of the electronic structure and the electrostatic potential, due to the iterative solution of these methods and the associated repetition of the expensive electronic structure calculation. This limits their application to chemical systems of small size. In this work, a new solvation model, ONIOM-EC-RISM, is presented which drastically reduces the computational cost of the previous EC-RISM method by introducing a multiscale approximation of the electronic structure. This opens up the possibility of applying the EC-RISM solvation model to larger chemical systems, such as protein systems. In addition to the relevant theoretical foundations of additive and subtractive multiscale approximations, required for the understanding of the ONIOM-EC-RISM model, the statistical-thermodynamical RISM solvation model and the methodologically related ONIOM-PCM solvation model, as well as the theory and technical implementation of the new model are presented in detail. In addition, it is shown how empirical corrections previously used for the EC-RISM reference model can be transferred to the ONIOM-EC-RISM context. Here, the size extrapolation limit of the ONIOM method is utilised, whereby corrections are obtained that are free of any partitioning errors. The resulting models are able to reproduce and partially exceed the pKa prediction quality of the EC-RISM reference model for the pKa dataset of the SAMPL6 challenge, a blind prediction challenge for thermodynamic quantities, while at the same time drastically reducing the overall cost of the EC-RISM method. In addition to the validation of the ONIOM-EC-RISM method on the basis of pKa values, it will be demonstrated how the ONIOM-EC-RISM model can additionally be used to predict chemical shifts of a pentapeptide system. In this context, a novel approach is presented that allows pH-dependent chemical shifts to be calculated directly from spectroscopic and pKa predictions. This opens up for the first time the possibility to directly model NMR titration experiments on the basis of the EC-RISM solvation model.
Description
Table of contents
Keywords
Theoretische Chemie, Solvatationsmodelle
Subjects based on RSWK
Theoretische Chemie, Solvatation, Integralgleichungsmethode