Quantum Accelerated Drug Discovery
The way pharmaceutical companies try to discover new drugs are similar to panning for gold. Wash a load of molecules slowly through a sieve, and pray for a sparkle of gold amidst the rubble. This early stage drug discovery costs pharma companies billions of dollars every year.
A London-based quantum start up is looking to change that needle-in-a-haystack approach. Vid Stojevic and Noor Shaker, two academics with backgrounds in tensor networks and machine learning, are co-founders of GTN. They met at London tech accelerator, Entrepreneur First, and felt that there was an opportunity to apply tensor networks to quantum data.
The pharmaceutical industry seemed like an obvious application, given the amount of data that the industry grapples with at a molecular level, as well as the characteristics of the data. The scientists both felt that the transition from classical methods to quantum-inspired algorithms would make sense for certain scenarios.
“We are building a platform that enables the generation of novel molecules,” says Vid. “We have a generative model, as well as a predictive model, specifically for binding affinity with respect to protein targets. The technology is built on classical foundations, but quantum-inspired. Therefore it’s easily portable to quantum machines. Once they exist!”
Although, pharma companies have huge computational resources, there are few machine learning experts within pharma, and the interdisciplinary expertise required for this type of modelling is rare to find. The impact, however, on the pharmaceutical industry could be transformational. The current process involving crude screening of huge databases of molecules means lots of good drug candidates are missed, and sometimes, poor candidates are selected. Replacing this with an instantaneous and accurate screen would be hugely beneficial in terms of speed, cost and accuracy.
Challenges abound. Scaling is tough for machine learning as many tools were built for individual experiments; molecular dynamics are also famously hard to replicate. In addition, pharma companies need convincing that external companies can provide a technology beyond their own internal capabilities. Traditionally of course, pharma companies are jealous guardians of their IP and reticent to let outsiders into their data.
“We have pilots in place with pharma companies, but beyond that, they would prefer to own the IP. We want to build more of a licensing model,” explains Vid.
What’s on the horizon for GTN?
“Hiring; we are up to 15 people, but need more! Also, we will be going to Series A funding. After the Series A, we’d like to have a milestone deal in place with a pharma partner. Finally, we’d like to have a hardware partnership in place. A lot of the really hard quantum questions will need customised hardware. That’s the only way we will truly prove pure quantum supremacy.”
The results that Vid, Noor and the team are demonstrating with their current technology will only multiply as the quantum computing foundations improve, ultimately helping to bring novel drugs to market cheaper and faster.
For more information email me directly on amit.das@alphaevents.com
Visit the GTN website at www.gtn.ai