Accelerating Drug Development and De-risking Drug Portfolios with AI+Quantum

November 2, 2023
Accelerating Drug Development and De-risking Drug Portfolios with AI+QuantumAccelerating Drug Development and De-risking Drug Portfolios with AI+Quantum

The field of biopharmaceuticals is evolving at a breakneck pace. But drug development itself remains stubbornly slow: to bring a new drug to market takes between 10 and 15 years. What’s more, failure rates hover around 78 percent, and promising candidates often fail late in the process. It’s also costly, ranging from $1.3B to $4B to develop a new drug. With so few treatments getting to market each year, even the most well-capitalized biopharma conglomerate cannot afford to pursue therapies that aren’t set up for success.

AQBioSim, our molecular simulation division, seeks to address these obstacles by leveraging AI and quantum technology to revolutionize drug development with end to end solutions. Although quantum computers are still years away from practical biopharma use, we’re able to simulate molecular interactions in silico using today’s advanced classical hardware. These simulations help researchers discover promising compounds faster, de-risk them for toxicity or other negative interactions, and identify other potential indications for which the compounds might be useful.

Because the interactions between proteins and drugs are governed by the principles of quantum mechanics, predicting the efficacy of a new compound requires a quantum-based approach. AI has been used in the past to run these kinds of simulations with mixed results. However, by utilizing high-powered, massively parallel GPUs and TPUs, AQBioSim can leverage classical and quantum-inspired algorithms paired with AI to simulate quantum-mechanical phenomena critical to identifying promising drugs. These quantum simulations are helping SandboxAQ’s customers achieve newfound accuracy in modeling the electronic structure of molecules, thus enabling more precise predictions and optimizations.

The exponentially-increasing power of AI is infused throughout our work at every stage. Our simulations themselves benefit from AI models trained on high-quality physics-based calculations to enhance their accuracy and speed. Our optimization and virtual screening algorithms use advanced Bayesian learning techniques to cover vastly more chemical space than would be possible with traditional methods alone. And our hypothesis generation and knowledge graph methods allow us to target our simulations at the level of scientific hypotheses, maximizing impact upon overall research programs. All this combines to bring molecules to experimental testing in a matter of months rather than years.

Physical Simulation With Today’s Hardware

At AQBioSim, we combine AI with physics- and quantum-information-based simulation to advance drug discovery. Our Absolute Free Energy Perturbation (AQ-FEP) software utilizes a proprietary computational method to predict binding affinities between molecules without need for reference data, making it uniquely beneficial for hit finding. By coupling with a unique active learning method, we can bring physics-based accuracy to the huge compound libraries used in modern drug discovery. Novel and traditionally undruggable candidate molecules can thus be accurately interrogated, without the need for a congeneric reference. Biopharma companies and other labs can then prioritize promising compounds, accelerate the drug discovery process, and make informed decisions.

We’re also harnessing the power of quantum entanglement. While arbitrary quantum states require quantum hardware to be simulated, the “area law” states that occur in drug discovery can in fact be classically simulated using novel advanced methods such as tensor networks. These innovative algorithms enable us to create efficient simulations of quantum systems, expediting complex computations that were once considered highly challenging.

One example is UCSF’s Institute for Neurodegenerative Diseases, which has used AQ-FEP to generate thousands of predictions with unprecedented speed and efficiency. 

“Our collaboration with SandboxAQ is accelerating lead optimization and is on track to take several years off our discovery timelines,” said Dr. Stanley B. Prusiner, Nobel laureate and director of the institute. “Our return on value from the SandboxAQ platform is substantial, as it is helping us get to clinical trials much faster with the best possible molecules for achieving a successful outcome.”

In addition to UCSF, we are working with Sanofi, Riboscience and other research labs to accelerate their development of life-saving drugs. 

If you would like to explore how AQBioSim can impact your life sciences company, please get in touch.

No items found.