Powering Drug Discovery with Quantum Mechanics Using HPC on AWS with QSimulate

QSimulate was founded in 2019 by Toru Shiozaki and Garnet Chan. As distinguished academics in the field of chemistry, the cofounders had years of experience leading research projects. “We were pushing the boundaries of quantum mechanics in academia, but that didn’t translate to real-world impact,” says Shiozaki, CEO and cofounder of QSimulate. “We decided to close the gap ourselves, and that’s why we started the company.” The founders hope that their company can help solve industrial-scale problems and make meaningful contributions to society with quantum technology.

For lead optimization in drug discovery, researchers typically rely on commercially available tools that use physics to simulate drug-protein interactions. Unlike quantum mechanical calculations, classical calculations are limited in the accuracy and drug-binding modalities that they can simulate. Quantum mechanical calculations can account for any modalities at a high accuracy; however, conventionally, they required powerful and extensive compute resources. Using on-premises supercomputers and traditional simulation tools, quantum-powered simulations could take years to run from start to finish. This has been deterring researchers from using quantum calculations due to both budgetary and time constraints.

QSimulate developed QUELO, a next-generation quantum mechanics platform, to accelerate lead optimization. To speed up its calculations, the company uses a breadth of AWS HPC resources to power its workloads. For example, QSimulate uses Amazon Elastic Compute Cloud (Amazon EC2), which offers secure and resizable compute capacity for virtually any workload. “We aim to establish ourselves as a powerhouse in computer-aided drug discovery,” says Shiozaki. “There’s a true sense of collaboration between AWS and QSimulate.”

QSimulate developed an initial version of QUELO that ran on CPU-based resources and could run quantum mechanics calculations in a matter of days. Since bringing QUELO to market, QSimulate has optimized its platform to increase the accuracy and speed of its calculations. QSimulate rearchitected its platform to run on GPU-based resources and designed proprietary mixed-precision algorithms, combining single-precision and double-precision floating-point formants. As a result, QUELO now runs efficiently on Amazon EC2 G6e Instances, which are also the most cost-efficient GPU-based instances for artificial intelligence inference and spatial-computing workloads. “We’ve found our sweet spot,” says Shiozaki. “Amazon EC2 G6e Instances help us run our calculations cost-effectively at peak performance.”

By combining mixed-precision algorithms and GPU-based resources, QSimulate has recently further reduced the runtime of its simulations, resulting in an aggregate speedup of close to 1,000x compared to existing technologies. This gives its customers the ability to run quantum calculations of protein drug complexes within milliseconds per snapshot, reducing total simulation times from months to hours. This translates to reducing customers’ compute costs by a factor of 100–1,000. “Simulations that used to be highly unrealistic because of limited compute resources can now be performed routinely,” says Shiozaki. “Even for use cases where conventional approaches are effective, we can provide better accuracy. Better accuracy means less experimentation, and that ultimately reduces the cost of drug development.”

To further accelerate adoption, QSimulate built QUELO on containers, facilitating fast cloud deployments. In seconds, QSimulate’s customers can deploy QUELO in the cloud or in their own IT environments. “QUELO is very user-friendly,” says Shiozaki. “Not only are we speeding up simulations, but also we’re helping our customers improve their operational efficiency, which means that they can support more research programs.” To manage its containers, QSimulate uses AWS ParallelCluster, a service for quickly building HPC compute environments on AWS.

For QSimulate, protecting its customers’ intellectual property is just as critical as compute performance. For an added layer of security, the company adopted AWS WAF, a service that protects web applications from common exploits. Data is stored in Amazon Relational Database Service (Amazon RDS), which makes it easy to manage relational databases optimized for total cost of ownership, and Amazon Simple Storage Service (Amazon S3), an object storage service built to retrieve any amount of data from anywhere. “We take advantage of many security offerings from AWS,” says Shiozaki. “All the things that come with those services, like encryption in transit, we enjoy using.”

QSimulate has doubled its customer base since releasing the latest version of QUELO in 2024. By using AWS, QSimulate is prepared for future growth. “We want to make a big impact on society and grow our company,” says Shiozaki. “As the company grows, we want to run our workload sustainably. One of the benefits of using AWS is that we can automatically make a contribution to carbon footprint reduction and sustainability.”

Its founders are also excited to explore untapped possibilities with quantum computing services like Amazon Braket, a service for accelerating quantum computing research. “Ultimately, what we want to do as a startup is create a completely computerized version of drug discovery that’s automated end-to-end,” says Shiozaki. “The elastic computing resources that AWS provides are crucial to helping us realize a quantum mechanical future.”