An anti-psoriasis drug owned by Takeda Pharmaceutical Co. in Japan could be the first drug to reach the market that has been created by artificial intelligence.

Before machine learning and artificial intelligence (AI), drug and vaccine development took years of trial and error. 

However, by combining those technological innovations, the formula for the new drug was finalized in six months: the AI was educated about the illness and the basics of biochemistry, then was turned loose to randomly combine molecules and predict the results.

When one potential outcome showed possibilities, the AI tinkered with it further until it either succeeded or proved impractical.

The result: Takeda’s new drug will enter third-stage human clinical trials later this year. If successful, it could yield annual sales of $3.7 billion, according to analysts at Jeffries.

AI can’t predict the range of quirky side effects drugs can bring or predict how effective a treatment will be; bench tests and animal trials are still necessary. However, AI can eliminate years’ worth of them by picking the best options to try in the first place.

As a result, in 2022 investors plowed $24.6 billion into companies using AI to invent new drugs, three times as much as in the preceding four years, according to research firm Deep Pharma Intelligence.

Over the next 10 years, AI will create as many as 50 new drugs that will garner at least $50 billion in additional sales for the industry, Morgan Stanley analysts estimate.

To grab a share of that windfall, AstraZeneca has partnered with U.K. firm  BenevolentAI and Illumina Inc. in San Diego to target new compounds.

Bayer, Roche, Takeda, and other pharma firms are working with Recursion Pharmaceuticals to tap the power of machine learning to find new formulas.

Takeda has more than 500 in-house experts in AI and machine learning and also is working with researchers at MIT. GSK has 160. Pfizer is collaborating with DeepMind and credits AI with speeding the development of Paxlovid, its anti-COVID pill by as much as 90 percent.

Pharma took notice of AI and machine learning in 2018, when Google’s DeepMind AI lab unveiled AlphaFold, which predicts the shape of proteins.

Determining proteins’ shapes is a fiendishly complex problem for humans, but essential in figuring out how a given disease works in the body. Once that’s known, scientists can begin to design protein shapes that neutralize the harmful one.

However, that process can take 10 years to bring a drug to market and 90 percent of bench experiments with promising compounds fail, according to the Scripps Research Translational Institute.

That’s a key reason why Big Pharma turned to AI and machine learning to speed the development of anti-COVID vaccines.

Scientists testing compounds in petri dishes might have to try tens of thousands of compounds to come up with a new drug, Jeb Keiper, CEO of Nimbus Therapeutics, told Bloomberg. 

Instead, AI and machine learning can present researchers with the 10 best ones to try in “wet chemistry” experiments. From those results, the AI can refine its research and make more accurate choices.

The process continues until the one compound that works best is chosen, tested, and finalized, Keiper explained.

TRENDPOST: Takeda’s anti-psoriasis medication may be the first AI-discovered drug to go commercial, but a host of others are right behind.

Recursion Pharmaceuticals has five on the way to treat cancer and rare illnesses. Exscientia has three, targeting cancer and inflammation. Insilico is well along with an AI-discovered drug that addresses a common form of pulmonary fibrosis.

Pharma has been transformed. Instead of scientists in white coats peering into beakers of chemicals for years, drugs increasingly will be discovered in a matter of months by computer and AI specialists.

That not only will ease human suffering and save lives—including those of lab animals used for testing—but also should make new drugs less expensive.