2x faster molecular simulation times on modern hardware
50% reduction in false positive drug structures
5-10 hours saved per week on manual data preparation
Ryvu Therapeutics scales molecular dynamics simulations, cutting processing time from weeks to hours and reducing false positives by 50%.
Ryvu Therapeutics scales molecular dynamics simulations, cutting processing time from weeks to hours and reducing false positives by 50%.
Mapping an incomprehensibly large chemical space
The search for next-generation cancer therapies is inherently a numbers game, but the scale is almost impossible to quantify. The chemical space of drug-like molecules is estimated at 1060 potential compounds, a volume far too massive for any laboratory to physically or computationally screen. For clinical stages to advance efficiently, clinical research groups must rely heavily on in silico methods to filter out dead ends and discover viable treatments.
"Computing technology has become essential in modern drug discovery," says Marcin Kowiel, director of data science and AI platforms at Ryvu Therapeutics. "Traditionally, prioritizing chemical compounds was a trial-and-error process that took years. Today, computational tools and machine learning compress that timeline at every stage and enable data-driven decisions.”
As a specialized biotechnology company focused on oncological therapies, Ryvu Therapeutics relies on complex methods like molecular docking, molecular dynamics, protein structure prediction, and physics-based free energy perturbation calculations. To make a real-world impact, its data science team needs to process hundreds of chemical proposals daily. Medicinal chemists require these rankings by the next morning to map out lab synthesis cycles without delay.
However, Ryvu Therapeutics’ fixed on-premises hardware pool created an operational bottleneck. A single molecular dynamics simulation can consume up to five hours of compute time. Multiply that by multiple parallel repetitions across hundreds of proposed compounds, and local pools quickly became overwhelmed. Turnarounds that should have been completed in hours often stretched into weeks, forcing scientists to rely on faster but less accurate screening techniques to preemptively cut down batch sizes.
To scale its operations and explore a broader chemical space, Ryvu Therapeutics needed to increase its compute infrastructure. The team chose Google Cloud for its cost predictability, data residency capabilities in the European region, and its seamless native integration with their existing Nextflow orchestration tool.
Computing technology has become essential in modern drug discovery. Traditionally, prioritizing chemical compounds was a trial-and-error process that took years. Today, computational tools and machine learning compress that timeline at every stage and enable data-driven decisions.
Marcin Kowiel
Director of Data Science and AI Platforms, Ryvu Therapeutics
Designing an automated computing funnel
Ryvu Therapeutics built a modern, automated in silico evaluation pipeline that uses Google Batch as the backend for its Nextflow orchestration. The data science team can now automatically spin up hundreds of GPU instances simultaneously to run complex molecular dynamics simulations.
Instead of spending hours manually scheduling experiments, the team submits compound proposals to the database, and the automated pipeline runs the simulations overnight. This scalable infrastructure generates terabytes of raw trajectory files, which are now securely stored and cached using Google Cloud Storage, completely eliminating the need for complex on-premises disk capacity planning.
Google Cloud has given us a reliable, scalable, and cost-manageable foundation for our most compute-intensive science, delivering a 2x simulation speedup, compressing iteration turnaround from weeks to days, and cutting false positives by 50%.
Marcin Kowiel
Director of Data Science and AI Platforms, Ryvu Therapeutics
This new system has shifted the rhythm of the company’s scientific workflow. Instead of waiting weeks for results, medicinal chemists receive automated, standardized reports every morning. By fully automating data preparation and reporting workflows, the data science team now saves five to 10 hours per week, freeing them to focus on developing advanced predictive machine learning models.
"Google Cloud has given us a reliable, scalable, and cost-manageable foundation for our most compute-intensive science, delivering a 2x simulation speedup, compressing iteration turnaround from weeks to days, and cutting false positives by 50%,” Kowiel says.
Enhancing drug quality and
preparing for an agentic future
The implementation of Google Cloud has transformed the pace of discovery at Ryvu Therapeutics. Running a standard campaign of 100 compounds requires the team to run three independent replication simulations per molecule to ensure scientific precision, equaling roughly 900 GPU hours of heavy parallel arithmetic. On Google Cloud, this computational load effortlessly completes in a single night, empowering medicinal chemists with timely charts every day.
“The automated pipeline delivers results in a standardized format with consistent naming, metadata, and reporting, so everyone sees the same information, presented the same way, every time,” Kowiel says. “That alone has removed a significant source of friction and error from our day-to-day work.”
The scale of Google Cloud also means scientists no longer have to choose between speed and data quality. Instead of filtering structures early with low-accuracy methods, the group runs high-fidelity molecular dynamics simulations across every structure proposal. Combining structural docking with data-rich dynamics models has reduced false positive compounds by 50%.
The automated pipeline delivers results in a standardized format with consistent naming, metadata, and reporting, so everyone sees the same information, presented the same way, every time. That alone has removed a significant source of friction and error from our day-to-day work.
Marcin Kowiel
Director of Data Science and AI Platforms, Ryvu Therapeutics
By filtering out inactive candidates before they reach lab synthesis, Ryvu Therapeutics has shaved months off the development life cycle and avoided massive experimental compound testing costs.
Automation has additionally freed the data science team from manual overhead. By fully automating data preparation and reporting workflows, the data science team now saves five to 10 hours per week, freeing them to focus on developing advanced predictive machine learning models. The team has reallocated this time to building custom machine learning frameworks, data augmentation, and prepping for an upcoming agentic transformation.
"Our success with Google Cloud has changed our internal culture, encouraging us to pursue larger, more ambitious project scopes," says Kowiel. "We are moving past previous limitations of what we can achieve. It allows us to go big."
Ryvu Therapeutics is a clinical-stage drug discovery and development company focusing on novel therapies that address emerging targets in oncology.
Industry: Healthcare and Life Sciences
Location: Poland
Products: Cloud Storage, Compute Engine, Gemini, Google Batch, Google Cloud
