Skip to main content
Premium Trial:

Request an Annual Quote

UK-Based Firm Taps Microsoft's Azure Cloud to Support Predictive Drug Discovery


UK-based drug discovery services firm Molplex is using Microsoft’s cloud infrastructure to build its computational compound screening business.

Molplex collaborated with Microsoft Research Connections to build its Clouds Against Disease service, which is based on Microsoft's Azure cloud platform and runs proprietary virtual screening algorithms.

David Leahy, Molplex’s CEO, explained that the company began developing the computational candidate screening technology about six or seven years ago.

“What we’ve been doing for the last couple of years is essentially validating the technology alongside real drug discovery projects and looking to raise investments, which we’ve managed to do now to expand the applications,” he told BioInform, though he did not disclose the nature of the investments.

The platform uses machine learning and multi-property optimization algorithms as well as search tools and chemical informatics systems to calculate the numerical properties of molecules.

Researchers at the firm will use the cloud platform to screen compounds for potential toxicity and biological efficacy to help partners discover potential drug compounds. Molplex promotes its approach as an advantage over traditional experiment-based drug discovery, which requires significant upfront work that would be wasted if the drug turns out to be toxic.

In one study looking at treatments for bacterial infections, Molplex used its cloud platform to screen more than 10,000 chemical structure and biological activity datasets. This generated about 750,000 predictive relationships between chemical structures and biological outcomes. The list was further pared down to 23,000 models covering 1,000 biological and physicochemical properties.

Once the models were built, Molplex used them to sift through a database of new chemicals to search for compounds that would be effective therapies.

“What we are doing requires an enormous amount of computing power in bursts,” Leahy said. “The challenge for us, particularly as we’ve been bootstrapping the business, is actually building a database and systems that can build and apply models across thousands of properties and hundreds of millions of structures.”

Leahy said that the company's cloud computing platform allows it to take a novel, "big data" approach to drug discovery.

“Traditionally you look for compounds that are active against a target, filter out those that are non-selective, and over time narrow down to compounds that can be absorbed and are non-toxic,” he explained.

Conversely, “what we do is build models across thousands of properties, including off-target [effects] and ADME and toxicity, and search for the drug straight on. That’s why we have this need for systems that can manage building models and optimizing across multiple properties,” he said.
“That means we are only pursuing in the laboratory those that look like they are going to be drug candidates from day one and we narrow that down very quickly and don’t spend money on things that we can’t predict.”

The Microsoft collaboration “has been critical,” Leahy continued. Having access to the cloud platform “clears away a number of barriers,” such as “the time that would be required [to run calculations] on a small number of physical servers,” he said.

Also, "if you are a small company then you [don’t] want to lay out $100,000 [to] $200,000 for servers that you are not going to use all the time,” Leahy said.

Molplex believes its computational approach can make drug discovery affordable for tropical diseases and niche disorders that are often a low priority for drug companies due to their limited commercial payoff.

“Our goal is to reduce the average cost of [drug discovery] from what’s currently about $30 million to $1 million or less,” Leahy said.

Although the company isn’t currently screening compounds for any pharma customers, it has used the platform to identify 10 drug candidates for bacterial infections, dermatology, oncology, and tropical diseases, Leahy said, although he declined to provide additional details.

These candidates were in part selected based on “our understanding of new research that gives us new insights into targets and mechanisms,” while others “are being defined in collaboration with people with commercial expertise,” he said.

Molplex expects to begin reaching out to pharmaceutical companies who might be interested in availing themselves of its screening service sometime next year, Leahy said.

“Our focus at the moment is to get our first internally driven pipeline up and underway,” he said. At that time, “we would look to have relationships with pharma companies and help them to define what we should be working on.”

Reaching Out

Microsoft’s partnership with Molplex began under the auspices of a program that was begun by Dennis Gannon, director of cloud research strategy at Microsoft Research Connections, and aimed to provide members of the external research community with access to its cloud infrastructure.

Under this program, Microsoft Research Connections has partnered with the National Science Foundation, the European Commission, and other groups to launch around 75 projects, Gannon told BioInform.

These include three life science projects conducted in partnership with the NSF, which provided free access to the Azure platform for two years. These projects were led by researchers at Virginia Tech; the University of North Carolina, Charlotte; and the J. Craig Venter Institute and were awarded a total of $1.2 million in grants.

Other projects include one with the University of Washington, Seattle, where researchers are using Azure to look at protein folding in Salmonella, Gannon said. Another project, in France, is looking at links between genetic patterns and brain anomalies, while a third project, in North Carolina, is using the cloud to run comparative genomics analysis.

In addition to providing the cloud infrastructure, Microsoft has developed a tool called Generic Worker — a worker-role implementation for Windows Azure — which is used to run computations in parallel on multiple processors, Gannon said.

Moving forward, Microsoft Research will take what “we learned from companies like Molplex and [from] building community-based data collections and services” and will apply these lessons to genomics projects, he said.

“There is a lot of high-quality genomic information that people need and use in the [Protein Databank] and the [Short Read Archive] and a number of these data collections … We are going to host a lot of that stuff on the cloud,” he said.

In addition, “[we want] to put the community tools that people really use out there on the cloud with really good web or app interfaces,” he said.

The goal for its next round of projects is “to reach out to the scientists that are not necessarily going to be … cloud programmers,” he said. Instead, “we want to reach the people who need to get their science done.”

For this next step, the group is partnering with researchers at Newcastle University who have built a software-as-a-service platform called e-science Central, a cloud-based data-analysis tool that Gannon said is currently being used in drug discovery but is being extended to work with genomics data.

Gannon also said that his group is working with scientists at the Technical University of Valencia, who are “gathering together different community tools” that will run on the Azure platform.

Microsoft Research is also seeking to forge partnerships with other science funding bodies such as the National Institutes of Health, Gannon said.