Artificial intelligence (AI) systems can predict protein structures, perform complex analyses, and help scientists move from idea to experiment.
The greatest opportunity for artificial intelligence will be at the intersection of biology and healthcare. From searching for new enzymes hidden in public databases to reading medical images for early diagnosis, artificial intelligence is already revolutionizing these fields.
Don’t believe me? In an anthropological survey of nearly 52,000 Americans, nearly half listed a cure for cancer or Alzheimer’s among their top hopes for artificial intelligence. Major AI labs see biology as the next frontier as well.
The clearest demonstration of what AI can currently do in biology is AlphaFold, whose creators shared half of the 2024 Nobel Prize in Chemistry. Proteins are chains of amino acids that form into three-dimensional shapes, which helps determine what they do. Scientists can read a protein’s chain of amino acids with relative ease. But sequencing alone doesn’t reveal the shape, which could take months or years to determine in the lab. AlphaFold predicts the likely shape of the sequence. Although it’s not perfect, it predicted the structures of more than 200 million proteins.
Two recently published papers show where AI may take biology next.
A paper in the journal Science describes Biomni, a research assistant in the field of artificial intelligence for biology and health research. The researcher can ask questions and use them to find and clean data, choose programs, write code, and analyze the results. In one published example, Bayoumni analyzed 458 wearable sensor files in less than an hour. The researchers estimated that a person would need about 60 hours to perform a similar task. She also designed the DNA fragments and laboratory steps for the CRISPR cloning experiment, which the scientists successfully completed.
Rubin, described in another Nature paper, is more of a semi-autonomous scientific collaborator. The researchers asked her to find possible treatments for dry age-related macular degeneration, a leading cause of blindness. It reviewed 551 research papers in 30 minutes and suggested it helps cells in the back of the eye get rid of waste. Human researchers then tested several of the candidates I proposed, and two of them improved this ability in cells. Robin then analyzed the results and suggested what to do next. She did not experiment or show that medications would preserve sight, but they helped move from the disease question to the experiment and then to the next hypothesis.
Read also: Scientifically: Do the first winners become adult winners?
We do not yet have a general artificial scientist who can pick an important question, devise an unusual experiment and see if the answer matters. But we are getting closer. Ginkgo Bioworks, which runs robotic laboratories, has gone a step further by connecting the robots to OpenAI’s AI system. The artificial intelligence suggested recipes for making the protein in the test mixture. The robots tested it and returned the results. Over the course of six runs and more than 36,000 interactions, the system was able to reduce the estimated cost per gram of protein by 40 percent compared to the standard used in the study.
Would this worry me if I were starting my career in biomedical sciences now? Absolutely not. Human experts will still be essential to identify important problems, verify work, and guide the direction of science. Ultimately, science is a human endeavor.
The scale of the problems we face, including those that the general public sees as having the greatest potential, will not be solved by AI alone.
Let’s take cancer. Cancer is not a single disease, but rather hundreds of diseases that arise as a result of different genetic changes in different tissues. Scientists have made tremendous progress without artificial intelligence. Some cancers can now be treated, and many cancers can be controlled for a much longer period than before. Cancer death rates overall have been declining for more than two decades.
But progress has been uneven. Sometimes scientists know exactly what causes cancer and are still struggling to stop it. For example, a faulty form of a protein known as KRAS can leave the growth signal permanently turned on. It looked like an obvious target, but the protein was smooth and compact, with almost nowhere for the drug to stick. The first drug to directly target a form of KRAS was approved only in 2021, nearly forty years after the target was identified. So far, this treatment has only worked for cancers that carry one specific change, and tumors could eventually find ways to overcome it.
Could AI have shortened this research? maybe. It may have helped screen more molecules or identify the hidden pocket sooner. But we will never know for sure. The final breakthrough still depends on decades of chemistry, experiments, and clinical trials. AI may speed up parts of this process. Cancer won’t be a problem with one easy solution.
Read also: Scientifically: Do the first winners become adult winners?
Alzheimer’s has been difficult for a different reason. For decades, scientists have focused on two abnormal proteins that accumulate in the brain. One, called amyloid, forms sticky plaques between brain cells. The other, Tao, forms entanglements within it. This led to a compelling theory that amyloid begins to accumulate first, tau damage follows, brain cells die, and memory begins to fail. The idea has gained support because rare inherited mutations that increase amyloid can cause Alzheimer’s disease unusually early in life.
But the common form of Alzheimer’s disease, which develops later in life, has proven to be more complex. Some people accumulate large amounts of amyloid without developing dementia. Many medications designed to remove it have failed. Newer treatments can remove amyloid and slightly slow cognitive and functional decline, but they do not stop or reverse the disease.
Amyloid may still be important. The problem may be that treatment begins too late, after the damage has spread. Or amyloid may be just one part of a disease that also includes tau, inflammation, blood vessels, and aging itself. This is why Alzheimer’s disease has been so difficult: Scientists may know some important pieces but not yet understand how they fit together. Trials are now testing whether removing amyloid before symptoms appear can delay dementia.
Artificial intelligence could one day help cure all diseases, said Demis Hassabis of Google DeepMind, one of the creators of AlphaFold. It’s an extraordinary ambition. But the history of cancer and Alzheimer’s disease suggests why faster prediction is insufficient. Biology is complex and messy in ways that neither machines nor humans can fully predict.
We will need scientists who understand digital and experimental biology. They must know what data to trust, what question is worth pursuing, and when an elegant answer doesn’t make sense in an actual cell or human body.
Anirban Mahapatra is a microbiologist, author, and scientific pioneer. His work spans the fields of microbes, medicine, artificial intelligence, and the institutions that shape science. His latest book is When Medications Don’t Work. The opinions expressed are personal
