Supercomputing Infrastructure

To understand biological systems and predict the impact of potential drug molecules, our scientists run accurate all-atom simulations at biologically meaningful timescales.

We start with a protein system comprised of approximately 50,000 atoms (including waters, cofactors, ions and all other biologically relevant molecules). We then perform millions of calculations to compute the interactions and forces between all atoms in the system. Based on the atomic forces, we update the position of all of the atoms in the system predicated on a femtosecond (one quadrillionth of a second) time step, then recalculate all of the interactions and update the positions a fraction of an Angstrom (one ten-billionth of a meter) again. When we do that 1000 times, we get a picosecond of biological time. When we do it a million times, we get a nanosecond, and when we do it a billion times, we get a microsecond, which is where a lot of the most interesting biology takes place (binding, allostery, loop movements, domain rearrangements, etc.). In short, we compute millions of atomic interactions billions of times for each potential drug molecule to get an accurate and statistically meaningful prediction of the system.

By the time we analyze millions of interactions, billions of times, we reach the point where we can accurately compute biologically meaningful quantities such as the binding energy and build a fundamental understanding how to conformationally modulate the target protein. From these simulations, we produce intuitive visualizations and data analyses that provide our chemists and biologists with actionable information to guide the next steps of drug design. To do this at scale requires hundreds of GPUs and thousands of CPUs for each of the Silicon Therapeutics drug discovery projects.

Purpose-Built High-Performance Computing

Silicon Therapeutics has built our high-performance computing environment to meet the company’s requirement for state-of-the-art computing power. We believe our dedicated, built-for-purpose, super-computing cluster provides more computing resources per project than any other biotechnology or pharmaceutical company. We continuously engage hundreds of GPUs aimed at each individual protein target, all running on the Silicon Therapeutics proprietary custom-coded software.

Silicon Therapeutics owns the entire spectrum of computational physics intellectual property, from the underlying computing infrastructure to the code and applications. We have invested in building these dedicated computing resources to ensure our scientists can pioneer exploration of this vast chemical space while meeting the demanding timelines required by our agile drug design and development process.

Data Science

We take a data-driven approach to drug discovery at Silicon Therapeutics. We employ advanced machine learning and artificial intelligence techniques when appropriate and traditional model building techniques at other times. Though we do not fit within the narrow definition of an artificial intelligence company because we are so much more, AI is a powerful component we use to process data from our integrated disciplines of biology, chemistry, biophysics and molecular simulations.