Application of large-scale quantum mechanical simulation to the development of future drug therapies

Lead Research Organisation: Newcastle University
Department Name: Sch of Natural Sciences & Env Sciences

Abstract

Rational computational design plays an increasingly important role in today's society, and is widely used in, for example, the construction and automotive industries to reduce costs associated with conventional experiments. If we are to apply the same principles to the design of pharmaceutical molecules, then it is necessary to be able to predict with high accuracy which of the multitude of molecules that we can potentially synthesise in the lab actually have therapeutic benefits. Ideally, the computer program would be able to perform this function using only established laws of physics, rather than relying on data input from experimental measurements. The modelling of atoms at this fundamental level is known as first principles simulation.

First principles simulations are used today by researchers in many industries, including microelectronics and renewable energy, to rapidly scan multitudes of hypothetical material compositions. Only once a set of materials matching the desired properties is discovered, does the costly process of manufacturing those materials in the lab begin. So why are the same first principles techniques not used to design new pharmaceutical molecules? The equations of quantum mechanics were written down and shown to describe the atomic-scale behaviour of materials with remarkable accuracy as early as the beginning of the twentieth century. Therefore, the answer is not a lack of physical understanding. Instead, it is largely a problem of the computational effort required to model the large numbers of atoms that are involved in interactions between a pharmaceutical molecule and its therapeutic target.

There are an unimaginable number of silicon atoms in typical modern electronic devices, but importantly the homogeneity of the structures means that the bulk material can be represented by just two atoms periodically repeated in 3D, and it is a relatively straightforward problem to computationally model the properties of this simple system. In contrast, biological systems are much more complex and often we need to simulate many thousands of atoms in order to accurately predict the relationships between the molecule's structure and its function. However, due to increases in computer power and, more importantly, fundamental advances in software design, first principles approaches can now access these biological systems with precisely the same accuracy that is used to study silicon.

Traditional approaches to computational drug discovery rely heavily on hundreds of model parameters that have been collected over many decades from experiments or computational analysis of small molecules. My idea is to dispense with these parameters and instead compute them directly from first principles quantum mechanical simulations of the biological therapeutic target, such as a protein that is implicated in disease. These new model parameters, rather than being generic, will be specific to the system under study and will thereby transform the accuracy of computational biomolecular modelling. The improved computational models will be used to scan hundreds of potential pharmaceutical molecules for therapeutic benefit, thus allowing us to rationally and rapidly design new therapeutic candidates. Medical researchers will be able to focus their design efforts on synthesising only the most promising molecules, thereby improving the likelihood of success in the early stages of pharmaceutical development and decreasing the cost of medicines to the patient. This concept will be put into practice in collaboration with the Northern Institute for Cancer Research at Newcastle University for the design of novel cancer therapies.

Planned Impact

Globally, between 1986 and 2000, human life expectancy increased by two years, and 40% of that increase is attributed to the use of new pharmaceuticals. The research and development of new medicines is an important sector of the UK economy. Medicines originating from UK companies accounted for 14% of the total sales of the world's top 100 selling drugs in 2014, and sales from the UK pharmaceutical industry are estimated to contribute a net £1 billion to the UK economy (source: www.abpi.org.uk).

This project will create a computational pre-screening method that will allow medicinal chemistry researchers to rapidly test hundreds of potential drug candidates on the computer and only synthesise in the lab a handful of molecules that are predicted to be the most successful, thus decreasing the experimental workload and costs. As a consequence, it is expected to have a significant economic impact by improving the efficiency and increasing the success rate of pharmaceutical research, and a societal impact by potentially bringing new medicines to market on a shorter time scale. To deliver this impact, the research team will design workflows that are accessible to medicinal chemists, and work with the Northern Institute for Cancer Research at Newcastle University and industrial partners at Astex Pharmaceuticals to demonstrate that this theoretical research is accurate enough to be translated into a protocol that can impact live drug discovery programmes. In turn, these project partners will obtain early access to the developed methods, thus enhancing the efficiency and competitiveness of UK pharmaceutical research and development.

The UK has a strong track record in quantum mechanical simulation and a particular competitive advantage in applying these methods to large, complex systems such as proteins. This is due in a large part to EPSRC funding of large-scale density functional theory codes, including ONETEP. This project will translate these strengths into the field of drug discovery by developing computational models of drug-target interactions based on large-scale quantum mechanical simulations using the ONETEP code, thus widening the scope and potential user base of this software. The open-source distribution of scripts and input/output files for force field design will facilitate the uptake of these methods by users in industry and academia.

The proposed project will provide a source of highly-trained Ph.D. and post-doctoral researchers with combined skills in scientific computing and medicinal chemistry. These skills will be highly sought after as the UK pharmaceutical industry continues to enhance the role of molecular modelling in its drug discovery processes.

This research will continue to have an impact beyond the timeframe of the current project. After 5 years, the designed workflows will be ready to be made available to the wider academic and industrial research communities for use in computer-aided drug design. After 10 years, lead optimisation efforts that incorporate the designed methods should demonstrate a measurable decrease in costs. In the longer term, new medicines will be designed with a substantial computational input from pre-screening methods such as the ones proposed here. Cost savings in pharmaceutical research will be passed onto the NHS, ultimately resulting in greater treatment options and improving the health and wellbeing of UK citizens.

Publications

10 25 50
 
Description Despite advances in drug design and structural biology, the optimisation of binding affinity between a candidate drug molecule and a therapeutic target remains a slow iterative process with a high degree of attrition, due for the most part to current inability to accurately model protein-ligand binding affinity. In this project we have developed methods to improve the accuracy of biomolecular atomistic modelling to deliver a technology that will allow us to go on to rationally design future drug therapies on the computer. In particular, we have written and released to the community two open source software toolkits for small molecule and protein force field parameterisation, which are designed to improve the accuracy and ease-of-use of methods to model molecular dynamics and interactions.

Four articles describing the preparation of parameters and application to protein-ligand binding have been published. Competitive accuracy with a widely-used biological force field is achieved, indicating that quantum mechanics derived force fields are approaching the accuracy required to provide guidance in prospective drug discovery campaigns. Dissemination of the methods is being pursued through the release of open source software (https://github.com/qubekit/QUBEKit) and invited invitations to speak at force field-related conferences and workshops.
Exploitation Route We have released the first versions of the software to the research community, and further development has been the subject of a successful application for follow-on funding (UKRI Future Leaders Fellowship). We are actively collaborating with Northern Institute for Cancer Research and are in regular contact with biotechnology project partners, with the goal of putting this technology to use for the efficient discovery of new medicines, and have joined the Open Force Field Initiative (https://openforcefield.org) to enable our software to be put to use for accurate force field design.
Sectors Chemicals,Digital/Communication/Information Technologies (including Software),Energy,Healthcare,Pharmaceuticals and Medical Biotechnology

URL https://blogs.ncl.ac.uk/danielcole/qube-force-field/
 
Description The ideas developed in this First Grant have led to follow-on funding (a UKRI Future Leaders Fellowship), which aims to transform the methods used to derive computational models of atomic-scale dynamics and interactions (force fields). As part of this work, I am now a co-investigator at the Open Force Field Initiative, which is a network of academic and industry researchers working together to advance science and infrastructure required for building the next generation of force fields. Through this route, and others, our ideas and software around force field design will influence the computational methods used in the pharmaceutical industry for computer-guided molecular design. Furthermore, the postdoctoral researcher employed on the First Grant is now employed by a UK pharmatech company, and is contributing to the economic competitiveness of the UK.
First Year Of Impact 2021
Sector Pharmaceuticals and Medical Biotechnology
Impact Types Economic

 
Description EPSRC Tier-2 High Performance Computing resources on CSD3 (40K GPU hours)
Amount £0 (GBP)
Organisation University of Cambridge 
Sector Academic/University
Country United Kingdom
Start 12/2018 
End 01/2020
 
Description UKRI Future Leaders Fellowship
Amount £1,314,427 (GBP)
Funding ID MR/T019654/1 
Organisation United Kingdom Research and Innovation 
Sector Public
Country United Kingdom
Start 09/2020 
End 09/2024
 
Title QUBEKit 
Description QUBEKit is a software toolkit for users to derive molecule-specific force fields for small organic molecules. It derives parameters directly from quantum mechanics in an automated manner, and writes simulation ready force field files, thus potentially improving the accuracy and ease-of-use of atomistic simulations in eg the computer-aided drug design field. 
Type Of Technology Software 
Year Produced 2018 
Open Source License? Yes  
Impact The publication describing the software has been cited 26 times (web of science). The software is now starting to be used outside my research group, eg to study ion trapping at graphene oxide surfaces (https://doi.org/10.1016/j.carbon.2020.12.032). 
URL https://doi.org/10.1021/acs.jcim.8b00767
 
Title QUBEMAKER 
Description QUBEMAKER is a software toolkit for preparing input files for atomistic simulations of proteins using our new force field. It is fully automated and derived from quantum mechanics. Release of the software is designed to improve the accessibility of our developed methods. 
Type Of Technology Software 
Year Produced 2018 
Impact The software has now been merged into our main QUBEKit software (https://github.com/qubekit/QUBEKit), and has been used to benchmark high-throughput free energy calculations (preprint: https://doi.org/10.26434/chemrxiv.13116878.v1). 
URL https://doi.org/10.1021/acsomega.9b01769
 
Description Lab twitter account 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Postgraduate students
Results and Impact Twitter account (@ColeGroupNCL) set up and reports research outcomes and links to further information (typically 1000+ engagements per tweet). ~650 followers include industry, recent science graduates and and potential employees. One Marie Curie fellowship application has been developed through a contact started through social media/website engagement.
Year(s) Of Engagement Activity 2018,2019,2020
 
Description Outreach blog post 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Public/other audiences
Results and Impact Worked with Lauren Nelson, a PhD student in my group, to write a blog highlighting some of our recent work in deriving force fields from quantum mechanics for applications in computer-aided drug design. Activity benefitted from my attendance at a Royal Society public outreach writing course (May 2019).
Year(s) Of Engagement Activity 2019
URL https://ashortscientist.wordpress.com/2019/10/18/our-group-research/
 
Description Postgraduate Conference - Making Connections 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Postgraduate students
Results and Impact On 5th July 2019, Ph D students Josh Horton and Chris Ringrose attended the first Faculty of Science, Agriculture & Engineering (SAgE) Postgraduate Conference (Newcastle University) "Making Connections". They gave a live demonstration of the predictive ability of our developed software (QUBEKit) at the Networking Science fair. Following interest on twitter, they also released all of the code used during the demo: https://github.com/cole-group/QUBEKit_examples/tree/master/Demonstration_day
Year(s) Of Engagement Activity 2019
URL https://blogs.ncl.ac.uk/danielcole/outreach/postgraduate-conference-making-connections/
 
Description Presentation and demonstration at ONETEP masterclass 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Postgraduate students
Results and Impact Around 30 participants from academia and industry attend the ONETEP masterclass, which aims to demonstrate to participants how to use the software in their own research projects. As part of this year's workshop, I gave a presentation on simulating biological systems with ONETEP and showed tutees how to use our software alongside ONETEP to derive biological force fields.
Year(s) Of Engagement Activity 2019
URL https://www.onetep.org/Main/MasterClass2019
 
Description Project website 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Postgraduate students
Results and Impact Website designed describing developed methods, aimed at academic researchers and industry. It has been advertised through all dissemination activities undertaken by the PI.
Year(s) Of Engagement Activity 2018,2019
URL https://blogs.ncl.ac.uk/danielcole/qube-force-field/