Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - Cloud-HTMD (A CLOUD APPLICATION PLATFORM FOR RATIONAL DRUG DISCOVERY USING HIGH THROUGHPUT MOLECULAR DYNAMICS)

Teaser

The Cloud-HTMD project aimed to reduce costs and improve the effectiveness of the drug discovery process. The general trend is that the search for new drugs, their validation and introduction into the market is getting more and more expensive (~2.000 M EUR per drug). Although...

Summary

The Cloud-HTMD project aimed to reduce costs and improve the effectiveness of the drug discovery process. The general trend is that the search for new drugs, their validation and introduction into the market is getting more and more expensive (~2.000 M EUR per drug). Although, most of the budget goes into clinical research, there is still a significant sum (40-50%) dedicated to the early discovery stage. This stage is critical as it shapes the candidate drug molecules to improve their efficacy and reduce their potential toxicity.

The clinical research can only be afforded by large pharmaceutical companies, while the early stage can be carried out by smaller entities, even academic laboratories. Nowadays, some experimental steps can be performed virtually (i.e. simulations), which drastically reduces the cost. However, those virtual experiments require expertise in preparing and carry out simulations, and, additionally, it needs expensive computational hardware. The idea behind our project was to allow any entity to predict their compounds characteristics by computational methods and only test few of them in clinical studies. If we could just reduce the number of compounds tested by a factor of two, we could reduce the cost of the early discovery stage by more than 30%, so it would save more than 300 million euros per drug (Acellera internal marketing analysis).

Thus, in this project, Acellera proposed a new integrated computational chemistry solution based on cloud computing. The objectives we wanted to achieve by recruiting an innovation associate were:
a) to develop a web-application to facilitate the use of HTMD in the cloud for biotechnology and pharmaceutical companies
b) to develop innovative medicinal chemistry protocols via a web-based and cloud-powered infrastructure.
c) to lower the existing barriers to access and use these powerful tools

We started by developing a solution to an elemental problem: parametrization of drug-like molecules, which is of outermost importance for any application in drug discovery (e.g. including small molecule or fragment screening, conformational analysis, cryptic pocket detection, etc). To obtain parameters, we needed to use quantum mechanic (QM) calculations, which are slow and expensive. There is a lack of methods to get robust parameters quickly and at low cost. So, the application developed by the innovation associate, contributed to answer this need.

Moreover, we demonstrated the use of cloud computing to accelerate the execution of biomolecular simulations. This effectively contributes to the reduction of the time invested and the cost associated to routinary steps of the drug discovery process. We estimate that the time dedicated to simulations is reduced by a factor of 100 and that the cost is considerably lower since it does not require the purchase and maintenance of expensive IT infrastructure. In the process, we have encountered difficulties to engage customers in using our tool because of security-related concerns due to the use of the cloud. Nevertheless, after extensive discussions with IT and legal departments, we managed to convince several customers, small to middle companies (than large pharmas) and nonprofit entities (e.g. universities).

As the web applications are run in the cloud with strict confidentiality, we cannot monitor specific usage or target systems, so it is hard for us to point out which is the pharmaceutical area that may benefit the most from this technology. We firmly believe that it can be useful in many areas, as molecular dynamics simulations offer an atomistic understanding (e.g protein/ligand interaction).

Work performed

The innovation associate first followed the training provided by Acellera on the use of HTMD. He also evaluated his strengths and weaknesses, and reformulated the project milestones for achieving the goals set.

As a next step, the associate developed a visual and user-friendly web-application to facilitate the execution of the protocols. This application, called parameterize, answers a critical issue when performing HTMD simulations. It automatizes the process of the key parameter fitting (i.e. atomic partial charges and dihedral angle parameters) from quantum mechanic data. Also, a set of molecules was benchmarked on the application, which represented typical fragments and drug-like molecules used in drug discovery.

Later, the associate adapted the fragment binding kinetics and allosteric binding pocket definition protocols to the cloud platform. The protocols were based on adaptive sampling and statistical analysis with the Markov state models. Fragments are commonly used in drug discovery to detect preliminary molecules having good affinity and offer the possibility to synthesize more easily more complex molecules based on them.

Finally, the protocols were benchmarked. We chose a challenging system, gPCR (3PBL) a class of proteins of special interest in drug discovery as they are involved in many diseases. It was proved that the protocol was able to reproduce the bind pose of a fragment (undisclosed as publication is in preparation) sharing the benzamide moiety. Unfortunately, the measurement of the binding kinetics was achieved due to insufficient efficiency of the sampling algorithm. The benchmark of allosteric pocket detection was delayed by the previous benchmark and results are still not available, but will be published in the near future.

The app is now available at www.playmolecule.org. The project results (parameterize App) have been disseminated in the conferences we participate in. Also in social media, we have announced the availability of the tool to our users and potential customers.

Final results

The project aimed to reduce the cost of drug discovery by developing a web application and protocols that make candidate drug molecule screening and optimization more streamlined and easier to access.

This was achieved through the deployment on the web (as it allows to use these tools anywhere in the world) and the use of cloud computing which avoids the investment in expensive computational hardware. It also democratizes the process of drug discovery allowing non-expert users to perform complicated tasks such as force field parameterization, enhanced sampling, and the Markov state protocols correctly and reliably from their web browser. For companies, this approach offers rapid scalability, as the cloud computing resources can be provisioned on need-to-use basis, while keeping the cost predictable and controllable.

We have several companies already interested in the results. However, due to security concerns and legal issues, the platform is deployed in the internal company infrastructure, rather than on the web. The feedback from our customers confirms the viability of the business model of using the cloud computing resources. To increase visibility, for academic users and non-profit organizations, we will provide access to the tools via the public website (www.playmolecule.org).

Furthermore, the tools are being used for internal projects of drug discovery at Acellera. This is contributing to the development of our internal know-how and will allow us to establish contacts and projects with third parties (pharmaceutical companies). Also, a scientific publication will be submitted soon, while another one is in preparation (handled by a third party).

Website & more info

More info: http://www.playmolecule.org.