Nvidia (NVDA.O)-backed SandboxAQ, an artificial intelligence business that broke out of Alphabet’s Google, opened a new tab on Wednesday and disclosed a wealth of data that it believes will help scientists better understand how pharmaceuticals bind to proteins, accelerating the development of novel medical therapies.
Helping scientists forecast whether a medication will attach to its target in the human body is the aim.
However, the data did not originate in a lab, even though it is supported by actual scientific experiments. Rather, the data was created using Nvidia’s chips by SandboxAQ, which has raised almost $1 billion in venture capital. SandboxAQ plans to feed the data back into AI models that scientists can use to quickly predict whether a small-molecule pharmaceutical will bind to the protein that researchers are targeting—a crucial question that needs to be addressed before a drug candidate can proceed.
Scientists can use the method to anticipate whether a drug molecule is likely to bind to the proteins involved in a biological process, such as the course of a disease, if the medicine’s purpose is to impede that process.
The method is a new area that blends advances in artificial intelligence with conventional scientific computing methods. Scientists have long developed equations that accurately predict how atoms will combine to form molecules in a variety of domains.
However, even with today’s fastest computers, the possible combinations become too great to compute by hand, even for relatively modest three-dimensional pharmaceutical compounds. In order to compute over 5.2 million new, “synthetic” three-dimensional molecules—molecules that haven’t been seen in the actual world but were computed using equations based on real-world data—SandboxAQ used the experimental data that was already available.
Artificial intelligence (AI) models that can forecast whether a novel medication molecule is likely to bind to the protein researchers are targeting in a fraction of the time it would take to compute it manually while maintaining accuracy can be trained using that synthetic data, which SandboxAQ is making publicly available. The data will be used to createSandboxAQ’s own AI models, which company hopes will produce outcomes comparable to conducting lab tests, but digitally.
News By Reuters
