The development of advanced simulation methodology to address challenges in the calculation of protein-ligand binding affinities
Lead Research Organisation:
University of Southampton
Department Name: Sch of Chemistry
Abstract
Research area:computational and theoretical chemistry.
Despite rigorous free energy calculations becoming increasingly widely used in industry and academia, there are still many situations where their agreement with experimental binding free energies is so poor as to be useless. The reasons for these failures are almost invariably associated with either inadequate configurational sampling during the molecular dynamics or Monte Carlo simulation, or an inadequate force field.
In this project, both areas will be addressed. First, a hybrid Monte Carlo (HMC) framework will be developed allowing a range of novel move types to be assessed. Hybrid Monte Carlo combines conventional molecular dynamics and Monte Carlo into a single approach, whereby molecular dynamics is used to sample the configuration space of a system, but with some bias introduced, and the effect of the bias is removed by the associated Monte Carlo test. In the first instance, we will take our grand canonical Monte Carlo method and couple it to molecular dynamics through HMC, but in the longer term we will look to introduce novel biasing moves including sampling from a biased velocity distribution or changes in ligand protonation. In this way we hope to be able to explore the configuration space of the protein-ligand complex more efficiently and hence yield more accurate binding affinities.
Second, to address force field deficiencies, our hybrid QM/MM methodology will be extended from calculating simple hydration free energies to protein-ligand binding free energies. While we have had considerable success with calculating hydration free energies, preliminary studies suggest that protein-ligand binding is a much harder problem, although the precise reasons for this are unclear. As a prototypical system, we will start by calculating the binding free energies of bound waters in the neuraminidase protein-ligand system, since obtaining a QM/MM ensemble of configurations is possible for such a small QM region (a single water molecule). Taking the lessons learned from this study, we will extend the simulations to real protein-ligand systems of pharmaceutical relevance.
Despite rigorous free energy calculations becoming increasingly widely used in industry and academia, there are still many situations where their agreement with experimental binding free energies is so poor as to be useless. The reasons for these failures are almost invariably associated with either inadequate configurational sampling during the molecular dynamics or Monte Carlo simulation, or an inadequate force field.
In this project, both areas will be addressed. First, a hybrid Monte Carlo (HMC) framework will be developed allowing a range of novel move types to be assessed. Hybrid Monte Carlo combines conventional molecular dynamics and Monte Carlo into a single approach, whereby molecular dynamics is used to sample the configuration space of a system, but with some bias introduced, and the effect of the bias is removed by the associated Monte Carlo test. In the first instance, we will take our grand canonical Monte Carlo method and couple it to molecular dynamics through HMC, but in the longer term we will look to introduce novel biasing moves including sampling from a biased velocity distribution or changes in ligand protonation. In this way we hope to be able to explore the configuration space of the protein-ligand complex more efficiently and hence yield more accurate binding affinities.
Second, to address force field deficiencies, our hybrid QM/MM methodology will be extended from calculating simple hydration free energies to protein-ligand binding free energies. While we have had considerable success with calculating hydration free energies, preliminary studies suggest that protein-ligand binding is a much harder problem, although the precise reasons for this are unclear. As a prototypical system, we will start by calculating the binding free energies of bound waters in the neuraminidase protein-ligand system, since obtaining a QM/MM ensemble of configurations is possible for such a small QM region (a single water molecule). Taking the lessons learned from this study, we will extend the simulations to real protein-ligand systems of pharmaceutical relevance.
Planned Impact
Modelling and simulation are playing an increasingly central role in all branches of science, both in Universities and in
industry, partly as a result of increasing computer power and partly through theoretical developments that provide more reliable models. Applications range from modelling chemical reactivity to simulation of hard, glassy, soft and biological materials; and modelling makes a decisive contribution to industry in areas such as drug design and delivery, modelling of reactivity and catalysis, and design of materials for opto-electronics and energy storage.
The UK (and all other leading economies) have recognised the need to invest heavily in High-Performance Computing to maintain economic competitiveness. We will deliver impact by training a generation of students equipped to develop new theoretical models; to provide software ready to leverage advantage from emerging computer architectures; and to pioneer the deployment of theory and modelling to new application domains in the chemical and allied sciences.
Our primary mechanisms for maximizing impact are:
(i) Through continual engagement, from the beginning, with industrial partners and academic colleagues to ensure clarity about their real training needs.
(ii) By ensuring that theory, as well as software and application, forms an integral part of training for all of our students: this is prioritised because the highest quality theoretical research in this area has led to game-changing impacts.
(iii) Through careful construction of a training model that emphasizes the importance of providing robust and sustainable software solutions for long-term application of modelling and simulation to real-world problems.
(iv) By an extensive programme of outreach activities, designed to ensure that the wider UK community derives direct and substantial benefit from our CDT, and that the mechanisms are in place to share best practice.
industry, partly as a result of increasing computer power and partly through theoretical developments that provide more reliable models. Applications range from modelling chemical reactivity to simulation of hard, glassy, soft and biological materials; and modelling makes a decisive contribution to industry in areas such as drug design and delivery, modelling of reactivity and catalysis, and design of materials for opto-electronics and energy storage.
The UK (and all other leading economies) have recognised the need to invest heavily in High-Performance Computing to maintain economic competitiveness. We will deliver impact by training a generation of students equipped to develop new theoretical models; to provide software ready to leverage advantage from emerging computer architectures; and to pioneer the deployment of theory and modelling to new application domains in the chemical and allied sciences.
Our primary mechanisms for maximizing impact are:
(i) Through continual engagement, from the beginning, with industrial partners and academic colleagues to ensure clarity about their real training needs.
(ii) By ensuring that theory, as well as software and application, forms an integral part of training for all of our students: this is prioritised because the highest quality theoretical research in this area has led to game-changing impacts.
(iii) Through careful construction of a training model that emphasizes the importance of providing robust and sustainable software solutions for long-term application of modelling and simulation to real-world problems.
(iv) By an extensive programme of outreach activities, designed to ensure that the wider UK community derives direct and substantial benefit from our CDT, and that the mechanisms are in place to share best practice.
Organisations
People |
ORCID iD |
| Jonathan Essex (Primary Supervisor) | |
| Khaled Maksoud (Student) |
| Description | Through this funding, a critical lens has been placed upon existing methodologies for simulating drug binding and interactions with target proteins that has uncovered important implications for the directions that computationally aided drug discovery should venture. Thus far, the efficiency of one of the most commonly used methods for accelerating protein dynamics in classical molecular dynamical simulations has been investigated by attempting to simulate the folding of a test miniprotein into its native state. This acted as a stress test for the method in question which highlighted key limitations in the ability of the method to consistently sampling a desired protein dynamical pathway and to output results in an operable and interpretative fashion. From these outputs, it is clear how such method may perform poorly in sampling the dynamics of complex systems that may be encountered as drug targets. This critical study was also necessary to inform the aspects of accelerated sampling algorithm development that must be addressed to allow for its operation upon modern drug targets. This indeed informed the development of a new algorithm that focuses upon extracting a molecular vibrational (infra-red) spectrum from a preliminary short simulation and leveraging this spectrum to directly accelerate the motion of the slowest modes of the system specifically using a digital filter applied to the spectrum. Such a method does not rely at all upon any a prioi knowledge of the system in question and allows for the spectrum to be obtained within a computational framework. However, the spectrum can be supplied from experimental studies and used directly, thus allowing for experimental findings to directly inform accelerated sampling |
| Exploitation Route | The new algorithm for accelerated sampling generated can be implemented into a computational workflow for the simulation of drug binding events on target proteins. This output is congruent both with the focus of academic research into structure-based drug discovery and the interests of pharmaceutical companies in developing and applying computer-aided drug discovery techniques to existing drug and protein target libraries |
| Sectors | Education,Healthcare,Pharmaceuticals and Medical Biotechnology |