Time filter

Source Type

Manchester, United Kingdom

Alkorta I.,Institute Quimica Medica IQM CSIC | Popelier P.L.A.,Manchester Institute of Biotechnology MIB | Popelier P.L.A.,University of Manchester
ChemPhysChem | Year: 2015

Remarkably simple yet effective linear free energy relationships were discovered between a single ab initio computed bond length in the gas phase and experimental pKa values in aqueous solution. The formation of these relationships is driven by chemical features such as functional groups, meta/para substitution and tautomerism. The high structural content of the ab initio bond length makes a given data set essentially divide itself into high correlation subsets (HCSs). Surprisingly, all molecules in a given high correlation subset share the same conformation in the gas phase. Here we show that accurate pKa values can be predicted from such HCSs. This is achieved within an accuracy of 0.2 pKa units for 5 drug molecules. © 2015 Wiley-VCH Verlag GmbH & Co. KGaA. Source

Popelier P.L.A.,Manchester Institute of Biotechnology MIB | Popelier P.L.A.,University of Manchester
Physica Scripta | Year: 2016

We are at the dawn of molecular simulations being carried out, literally, by atoms endowed by knowledge of how to behave quantum mechanically in the vicinity of other atoms. The 'next-next-generation' force field that aims to achieve this is called QCTFF, for now, although a more pronounceable name will be suggested in the conclusion. Classical force fields such as AMBER mimic the interatomic energy experienced by atoms during a molecular simulation, with simple expressions capturing a relationship between energy and nuclear position. Such force fields neither see the electron density nor exchange-delocalization itself, or exact electrostatic interaction; they only contain simple equations and elementary parameters such as point charges to imitate the energies between atoms. Next-generation force fields, such as AMOEBA, go further and make the electrostatics more accurate by introducing multipole moments and dipolar polarization. However, QCTFF goes even further and abolishes all traditional force field expressions (e.g. Hooke's law and extensions, Lennard-Jones) in favor of atomistic kriging models. These machine learning models learn how fundamental energy quantities, as well as high-rank multipole moments, all associated with an atom of interest, vary with the precise positions of atomic neighbors. As a result, all structural phenomena can be rapidly calculated as an interplay of intra-atomic energy, exchange-delocalization energy, electrostatic energy and dynamic correlation energy. The final QCTFF force field will generate a wealth of localized quantum information while being faster than a Car-Parrinello simulation (which does not generate local information). Isn't it enough to see that a garden is beautiful without having to believe that there are fairies at the bottom of it too? (Douglas Adams). © 2016 The Royal Swedish Academy of Sciences. Source

Hughes T.J.,University of Manchester | Kandathil S.M.,University of Manchester | Popelier P.L.A.,Manchester Institute of Biotechnology MIB
Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy | Year: 2014

As intermolecular interactions such as the hydrogen bond are electrostatic in origin, rigorous treatment of this term within force field methodologies should be mandatory. We present a method able of accurately reproducing such interactions for seven van der Waals complexes. It uses atomic multipole moments up to hexadecupole moment mapped to the positions of the nuclear coordinates by the machine learning method kriging. Models were built at three levels of theory: HF/6-31G∗∗, B3LYP/aug-cc-pVDZ and M06-2X/aug-cc-pVDZ. The quality of the kriging models was measured by their ability to predict the electrostatic interaction energy between atoms in external test examples for which the true energies are known. At all levels of theory, >90% of test cases for small van der Waals complexes were predicted within 1 kJ mol-1, decreasing to 60-70% of test cases for larger base pair complexes. Models built on moments obtained at B3LYP and M06-2X level generally outperformed those at HF level. For all systems the individual interactions were predicted with a mean unsigned error of less than 1 kJ mol-1. © 2014 Elsevier B.V. All rights reserved. Source

Fletcher T.L.,University of Manchester | Davie S.J.,University of Manchester | Popelier P.L.A.,Manchester Institute of Biotechnology MIB
Journal of Chemical Theory and Computation | Year: 2014

Present computing power enables novel ways of modeling polarization. Here we show that the machine learning method kriging accurately captures the way the electron density of a topological atom responds to a change in the positions of the surrounding atoms. The success of this method is demonstrated on the four aromatic amino acids histidine, phenylalanine, tryptophan, and tyrosine. A new technique of varying training set sizes to vastly reduce training times while maintaining accuracy is described and applied to each amino acid. Each amino acid has its geometry distorted via normal modes of vibration over all local energy minima in the Ramachandran map. These geometries are then used to train the kriging models. Total electrostatic energies predicted by the kriging models for previously unseen geometries are compared to the true energies, yielding mean absolute errors of 2.9, 5.1, 4.2, and 2.8 kJ mol-1for histidine, phenylalanine, tryptophan, and tyrosine, respectively. © 2014 American Chemical Society. Source

Griffiths M.Z.,Manchester Institute of Biotechnology MIB | Griffiths M.Z.,University of Manchester | Alkorta I.,Institute Quimica Medica IQM CSIC | Popelier P.L.A.,Manchester Institute of Biotechnology MIB | Popelier P.L.A.,University of Manchester
Molecular Informatics | Year: 2013

Here we applied a novel method1a to predict pKa values of the guanidine functional group, which is a notoriously difficult. This method, which was developed in our lab, uses only one ab initio bond length obtained at a low level of theory. The method is shown to work for drug molecules, delivers prediction errors of less than 0.5 log units, successfully treats tautomerisation in close relation with experiment, and demonstrates strong correlations with only a few data points. The high structural content of the ab initio bond length makes a given data set essentially divide itself into high correlation subsets. One then observes that molecules within a subset possess a common substructure. Each high correlation subset exists in its own region of chemical space. The high correlation subset method is explored with respect to this position in chemical space, in particular tautomerisation. The proposed method is able to distinguish between different tautomeric forms and the preferred tautomeric form emerges naturally, in agreement with experiment. Copyright © 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. Source

Discover hidden collaborations