Medicine is AI’s real test, and India has a historic opportunity: Siddhartha Mukherjee

Artificial intelligence (AI) should be judged by whether it makes medicines faster, cheaper and better, oncologist and author Siddhartha Mukherjee said on Saturday, warning that if it does not, "we have failed the AI revolution".

Mukherjee, co-founder and chief executive of the drug-discovery company Manas AI, opened his talk at the Hindustan Times Leadership Summit by noting the widespread concern over AI’s rapid rise. Its capacity to produce transformative medicine, he said, was the one feature that could offset that concern, and if AI failed there, doubts about its value would return. He returned to the point at the end of his talk: “If it’s not true that medicines become faster, cheaper, better, then we have failed the AI revolution."

Medicine has been moving the wrong way, he said. “The number of medicines approved over time has actually been decreasing… and the cost of each medicine has been increasing."

Drug researchers call this pattern Eroom's Law, Moore's Law spelt backwards.

If AI can reverse it, Mukherjee said, "we will truly have mastered the capacity of bringing medicines to vast numbers of people".

He said India could help lead that change, and called it the main message of his talk. "Of all countries in the world, China and I truly believe India, with appropriate regulatory and technological reform, can reverse the trend," he said. "We have a historic opportunity in this country."

Mukherjee explained that every medicine binds to a protein in the body the way a key fits a lock. He said aspirin, Lipitor and Tylenol all work this way.

Mukherjee said trying every possible key is out of the question. He added that small-molecule drugs number 10 to the power of 24. Mukherjee said that no amount of power in the current system can possibly test 10 to the 24th keys against one single target protein.

Mukherjee said modern drug design instead identifies the target, finds its "keyhole," and builds a chemical from scratch to fit it.

He added that Manas does this with a combination of artificial intelligence, which is a neural system, and what they started calling medicinal intelligence, meaning the rules of physics and chemistry.

Mukherjee said that for daraxonrasib, a pancreatic cancer drug, a human researcher took about 10 years to find the keyhole, but their algorithm took two hours to find it. "That is the kind of speed that we get, the kind of compression of speed that we get just from this module, just from the keyhole finding."

Mukherjee said Manas sends 50 to 100 proposed compounds every three weeks to Chinese chemical companies, which test them and return data in about three weeks. He added that the model learns from each batch before the next is made until it reaches a medicine. Mukherjee called this the design-make-test cycle, and one of the fastest they have ever been able to achieve.

He gave an Indian example of what AI could deliver. About three years ago, he said, he launched Immuneel, a cell-therapy company in Bengaluru, in collaboration with Kiran Mazumdar-Shaw. In cell therapy, doctors extract a patient’s T cells, “weaponise” them in a laboratory to kill the cancer, grow them for 10 to 12 days and infuse them back. Mukherjee said the treatment, used against leukaemia and lymphoma and now being extended to myeloma, is potentially curative and was not available anywhere in India before.