SandboxAQ has been quietly working for some time to develop drug discovery solutions that can leverage hybrid CPU/TPU (Tensor Processing Unit) configurations and AI technologies to deliver meaningful outcomes for drug discovery. They have now created a new division, called AQBioSim, to bring it to market. As well as having computer and quantum scientists to work on these problems, the company has hired its own medicinal chemistry team to ensure they can provide a focused solution perspective.. The company has cooperated with University of San Francisco Institute of Neurodegenerative Diseases to test their technology as well as initial customers included AstraZeneca And Sanofi.
As we have discussed many times on this page, there is great potential to significantly improve drug discovery by applying computational simulation techniques to estimate the potential performance of drug candidates. in-silico rather it is physically testing compounds in test tubes or on animal or human subjects. Recent estimates are that it takes about 13 years and $2.5 billion dollars for a drug company to successfully deploy a drug from concept to FDA approval. And so far, approximately 70% of candidate drugs come out because they don’t work or have bad side effects.
There are programs in both the classical and quantum industries to change all that in hopes of reducing costs and time to market for new drugs or ingredients by an order of magnitude or more. This is not to say that computational chemistry simulations will completely eliminate field experiments; they will always be needed. But using simulation can make a big difference by greatly narrowing the number of drug candidates that need to be field tested and resulting in major time and money savings.
Computational chemical simulation has been studied for a long time with classical computers. However, the problem was that the number of computations became so large that it was no longer possible for a standard classical computer to give a correct answer. The industry has also provided various algorithms that use estimates rather than arriving at a precise answer. This greatly reduces computation time, but in many cases this approach does not provide suitable accuracy. The situation is changing due to the continuous improvements in artificial intelligence, quantum computing and new classical computing architectures such as GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units). The performance of these GPU and TPU chips has improved greatly over the last few years and their architecture is better suited than standard microprocessors for linear algebra based algorithms used in computational chemistry.
While SandboxAQ’s current offering does not use actual quantum computers, they hope to migrate future versions of their drug discovery technology to quantum computers when larger, higher-performance quantum machines become available. One advantage of SandboxAQ’s approach is that they are not tied to one particular hardware or cloud vendor. So they can take advantage of the best generation when selecting the appropriate hardware platform.
For more information on SandboxAQ’s announcement of their new AQBioSim division, you can access the press release available on their website. Here. Also, Jack Hidary, CEO of SandboxAQ, was recently interviewed on CNBC television about this endeavor and you can check out the video of his interview Here. Additionally, the researchers at SandboxAQ have written several technical papers explaining in greater detail how they were able to use the TPU for the complex linear algebra calculations used for this endeavour. You can see the papers Here And Here.
June 22, 2023