AI has long been touted as the future of drug discovery. Just plug in the AI and pump out the miracle cures. Simple, right?

Not exactly and not just yet, according to researchers led by Khalifa University’s Andreas Bender.

The team took a step back to evaluate the real-world factors that complicate the process of AI-assisted drug discovery.

In their paper published in Nature Reviews Drug Discovery, Bender and his colleagues argue that AI in drug discovery needs to address some challenges better, including:

• Researchers need to consider better how the AI will eventually be used in clinical settings when they are developing it.
• AI needs biological and medical data that is predictive for the question the model aims to answer.
• Researchers need to clearly define the problem the AI is supposed to solve, which then can lead to models that are properly equipped to address real-world situations.

“When I worked at AstraZeneca, people always talked about the right drug, in the right patient, dosed in the right way,” Bender tells KUSTReview.com. “We can, and need to, translate this thinking also to computational models and AI in drug discovery – which then means the right data, in the right model, validated and used in the right way.”

More like this: AI viruses join the fight against bacteria