The News
Artificial intelligence will “revolutionize” the process by which drugs are discovered, chip giant Nvidia said this week, after unveiling a pilot project for Danish drugmaker Novo Nordisk to use its new AI-powered supercomputer.
“Computer-aided drug discovery, I think that’s going to revolutionize the industry,” Nvidia CEO Jensen Huang said at a launch event for the computer, which researchers will use to train an AI model to facilitate vaccine design and analyze disease mutations, Bloomberg reported.
Machine learning has excited the pharma industry with its potential — for example, by scanning millions of possibilities to assess the effectiveness of shifting a drug to treat a different disease than first intended, replacing months of lab work.
One major breakthrough was Google DeepMind’s AlphaFold software, which predicts the structure and interactions of molecules — a previously complex, time-consuming process. Its inventors were awarded the Nobel Prize in Chemistry this month.
SIGNALS
AI’s main advantages for drug discovery are speed and cost-cutting
The key advantages that AI brings to drug discovery are speed and cost-effectiveness. Without the technology, a new drug typically costs in excess of $1 billion to develop, and take more than a decade to reach the market, according to a 2020 study published in the journal JAMA. AI has the potential to reduce that timeline significantly, using machine learning models that sort through millions of pieces of data, predict how certain compounds may affect the body, and discard unsuccessful ones at the computer stage — currently a painstaking process carried out by humans in labs, the MIT Technology Review noted. “It’s already doing a lot of the steps that we used to do by hand,” a researcher told the outlet.
The impact of AI-aided drug discovery is still unclear
A key test for pharma companies relying on AI will be understanding whether drugs developed this way are as effective for patients, Bloomberg noted. While some human trials for AI-aided drugs are underway, the true impact of the technology on the pharmaceutical industry won’t be assessed until more data is available. “Sometimes when you open the door, you find out that there’s actually nothing behind it,” a physicist and founder of a biotech machine learning company told the outlet, adding that AI is “not a panacea that can solve everything.”
Use of AI in pharma is fraught with ethical concerns
AI use in drug development is still in its early stages, and there are ethical factors to consider, Stanford Medicine’s Scope blog noted — the technology’s well-documented algorithmic bias related to gender, ethnicity, and sexual orientation, for example, the unknown consequences of using AI without human mediation, and the difficulty of reproducing clinical trial results when AI models are used incorrectly, which can render them virtually useless. Privacy is also a major concern, because AI systems rely on large amounts of data that can be misused, and health care data is particularly sensitive. Considerations around how to reduce the bias in algorithms and how to preserve privacy should be at the core of the development of the technology, a study in the journal Pharmaceuticals argued.
