Schrödinger: Expediting drug discovery using the cloud
About Schrödinger
Headquartered in New York City, Schrödinger is a computational chemistry and chemoinformatics company that specializes in accelerating drug discovery and materials science using physics-based computational modeling.
Tell us your challenge. We're here to help.
Contact usSchrödinger uses physics-based simulations carried out on the cloud in combination with machine learning to accelerate the discovery of new medicines and materials.
Transforming the process of drug discovery through better candidate selection
Drug discovery is an arduous process—expensive and time-consuming, with long odds for success. Traditionally, chemists and pharmaceutical scientists have chosen which candidate drugs to synthesize and test based on their knowledge of existing pharmaceuticals, research about the disease they are attempting to treat, and their intuition about what seems most likely to work. The scientists at Schrödinger aim to disrupt that paradigm. Their vision: to transform the way researchers discover new therapeutics and materials through the power of computational modeling.
Thirty years after the company was founded, Schrödinger’s scientists have a lot to celebrate. The company’s advanced computational platform expands the scope of exploration and enables today’s drug hunters to evaluate billions of therapeutic candidates, and to identify those with the most optimal properties. By testing drug candidates in silico, Schrödinger can quickly and efficiently identify which of the countless options are worth synthesizing and taking into the lab for further testing. This process reduces the time and expense of getting a new therapeutic candidate through the discovery phase. Just as important, the computational platform improves the likelihood that the therapeutic candidates will succeed in clinical trials, because they will be rigorously tested for everything from solubility to selectivity in the computer simulations.
The ultimate goal: To get important new medicines to patients more quickly and with less up-front expense.
Schrödinger’s rigorous focus on physics-based computation requires a huge amount of computing power. The company built data centers with massive capacity but came to realize these on-premises data centers were not enough. Drug discovery projects tend to run in bursts; Schrödinger’s teams often needed huge amounts of computing power for just a few days each month. Rather than build more capacity and let it sit idle for long stretches of time, Schrödinger decided to migrate its groundbreaking drug discovery work to the cloud.
In evaluating options, Schrödinger’s team looked for a cloud provider that was just as committed to advancing life sciences and drug discovery as it was. The company chose Google Cloud, which is known for the strength of its network and security posture.
Computational power to fuel groundbreaking drug discovery
With Google Cloud, Schrödinger has access to a near-infinite volume of processing power on demand, allowing it to run simulations that required bursts of on-demand compute power that were hard to supply with its on-premises data centers. With high-performance computing resources on Google Cloud, the Schrödinger team can explore a larger chemical space and model vastly more compounds, improving the odds of identifying a promising and novel therapeutic candidate for a given disease.
The company also is able to run simulations without concern over network stability. As Robert Abel, Ph.D., Executive Vice President of Science at Schrödinger, explains, “Being able to fit everything on one GPU card advanced the stability and effectiveness of these simulations. Cloud computing made that possible.” When a simulation required an especially computationally intensive burst, the Cloud team made additional resources available. Schrödinger CEO Ramy Farid, Ph.D., says, “Availability of GPUs is very important. With Google Cloud, a request to provision 50,000 or 100,000 GPUs wasn’t a barrier.”
This is why finding the right cloud partner was so important to Schrödinger. Between the computationally intensive nature of these simulations and the need for total network stability to run the simulations to completion, traditional processors simply would not work. Cloud computing was mandatory if Schrödinger wanted to use its computational platform to discover new drugs or materials. “The physics-based simulations require about a GPU day per molecule,” explains Dr. Abel. “That's the equivalent of about 100 to 200 CPU days per molecule. That’s not so bad to commit per molecule, but for simulations of thousands of molecules interacting with each other, you need reliable network connections between all of those CPUs. Transient network instabilities would crash the jobs.”
Accurately modeling the dynamic interactions between even one potential drug molecule and its protein target in a realistic biological environment already requires a lot of processor power. But in order to identify the most promising candidate, you must test thousands, millions, or even billions of possible compounds. And to ensure selectivity, you have to test against multiple protein targets. “To impact a project, you need to spend the weekend running 100,000 processors at a time. Then, once you’ve identified the most promising molecules, you go to the lab and make compounds to test them further. You can't keep a 100,000-processor cluster busy 100 percent of the time. That’s what makes this a perfect use case for the cloud. You can burst processor capacity without investing in a cluster that sits idle a lot of the time,” says Dr. Farid.
“What stood out was the way the Google Cloud team shared our belief in the potential of combining the right platform with the right technology. This partnership could lead to a broader adoption of these methods by the pharmaceutical industry—and could potentially transform the way all of pharma is doing drug discovery."
—Dr. Ramy Farid, CEO, SchrödingerUsing Google Cloud also helped Schrödinger manage its bottom line. “Of course, cost is an important factor,” says Dr. Farid. “We're using these resources on a massive scale, and those expenses start to add up.” Schrödinger took advantage of the ability to customize the type of VM that the GPUs use, increasing flexibility and cost efficiency.
The Schrödinger team believes running its advanced computational platform on the cloud could have a ripple effect even beyond the drug discovery projects it is conducting internally to include work with biopharma collaborators. “What stood out was the way the Google Cloud team shared our belief in the potential of combining the right platform with the right technology. This partnership could lead to a broader adoption of these methods by the pharmaceutical industry—and could potentially transform the way all of pharma is doing drug discovery,” Dr. Farid says.
Expedited drug discovery when it matters most
In addition to ongoing development of its computational chemistry platform, Schrödinger also has an internal drug discovery team. The Schrödinger team is advancing five internal programs to discover new drugs for specific targets in oncology, in addition to numerous programs that the team is working on together with biotech and pharma collaborators around the globe.
In one of the more exciting collaborations, Schrödinger has embarked on a large-scale philanthropic effort to identify novel antiviral therapeutics that could be used to treat patients infected with SARS-Cov-2. Working in collaboration with Takeda, Novartis, Gilead Sciences, WuXi AppTec, and Google Cloud, the company hopes to accelerate drug discovery using its computational platform. Schrödinger’s participation during the COVID-19 pandemic highlights the need for efficient drug discovery methods that accelerate time-to-market. “COVID-19 really brings it into focus. We have a huge, unmet clinical need for better, more effective drug therapies. Our methods increase the probability of finding a molecule that can be advanced into clinical studies. We also expect to find that molecule more rapidly,” Dr. Abel says. “Traditional drug discovery projects, when successful, take about five to six years to get into clinical trials. By rapidly identifying good molecules with properties that justify advancement into clinical studies, we believe our platform may yield a faster timeline, potentially in the range of two to three years to the clinic.”
Schrödinger’s methods may be particularly well suited to the challenge of finding therapeutically effective treatments for novel infectious agents like SARS-CoV-2. “These methods allow for rapid hypothesis testing,” explains Dr. Abel. “There are a number of unknowns around the biology and chemistry of coronavirus. We believe that enabling the identification of small molecules that engage different targets which are expected to be therapeutically relevant will allow us to expedite the discovery of an effective small-molecule drug therapy for COVID-19.”
Dr. Abel concludes, “Given the difficulties we're all facing, expanding the world’s drug discovery toolkit will be useful for the present outbreak and beyond.”
Tell us your challenge. We're here to help.
Contact us
About Schrödinger
Headquartered in New York City, Schrödinger is a computational chemistry and chemoinformatics company that specializes in accelerating drug discovery and materials science using physics-based computational modeling.