BIOLOGY + ARTIFICIAL INTELLIGENCE THE YOTTABIT ERA
What if we could design medicines for diseases we can't treat today?
We are beginning to move from searching for useful molecules to designing biological tools for particular targets. That could change how the search for new treatments begins.
The whole story.
In one minute.
ONE STORY.
- 01
Artificial intelligence has helped scientists predict the shapes of more than 200 million proteins. These tiny biological structures do much of the work inside living things, and understanding their shapes gives researchers a far better starting point for investigating disease.
- 02
Now researchers are moving beyond predicting existing proteins toward designing new ones. In laboratory experiments, scientists have created computer-designed antibodies intended to attach to carefully selected biological targets, then tested whether those molecules actually worked as designed.
- 03
That could change the early search for medicines. Instead of testing vast numbers of possibilities with little guidance, research teams could increasingly use computing to select promising molecular designs and concentrate physical experiments on the strongest candidates.
- 04
If this approach improves, it could open new research paths for cancers, rare diseases and illnesses that have resisted conventional drug discovery. But binding to a target in a laboratory is only an early step; effectiveness and safety in patients must still be established.
- 05
The extraordinary possibility is a new kind of medical engineering. Artificial intelligence, genetic knowledge and automated laboratories could eventually work together to design, test and improve biological tools much faster than traditional discovery alone.
predicted protein structures now freely accessible to scientists — not 200 million medicines.
It's more than a breakthrough.
It's a different future.
Imagine a family receiving a diagnosis for a condition that medicine has struggled to treat. The doctors understand the disease, but there is no reliable drug that attacks the biological process behind it. The family enters the familiar world of waiting, trying available options and hoping that new science arrives in time.
Now imagine researchers tackling that problem from a different direction. They begin with a protein involved in the illness, examine its shape and ask whether they can design a molecule to interact with one specific part of it. Instead of searching almost blindly for a useful drug, they use powerful computers to propose designs and laboratory equipment to test them.
That does not mean a cure can be produced on demand. It means something potentially extraordinary is changing about the way the search for medicines can begin.
We are learning to read the machinery of life
Proteins are essential working parts of every living thing. They carry signals between cells, help fight infection and perform the chemical tasks that keep bodies functioning. A protein’s three-dimensional shape has a great deal to do with what it can do, which is why scientists have spent decades trying to understand how proteins fold.
Determining a structure in a laboratory can be difficult and expensive. Artificial intelligence has made it possible to predict many protein shapes from the chemical sequence that makes them up. The AlphaFold Protein Structure Database now makes more than 200 million such predictions available to researchers around the world. These are predictions, not 200 million experimentally confirmed structures, but the scale of accessible information is remarkable.
Think about the difference between trying to repair an unfamiliar machine without knowing its parts and having a detailed guide to how many of those parts might fit together. The guide is not perfect and it cannot tell you everything about how the machine operates. But it can help you ask better questions and decide which experiments to perform next.
The next step is more astonishing: design a new part
Predicting the shape of a protein that already exists is one achievement. Designing a new biological molecule with a particular task in mind is another. Researchers are beginning to connect these capabilities, using artificial intelligence to create possible proteins and antibodies that might interact with chosen molecular targets.
A study published in Nature in November 2025 demonstrated computer-designed antibodies that bound to selected areas of proteins associated with influenza and a toxin produced by the bacterium Clostridioides difficile. The researchers tested these molecules in the laboratory and used experimental measurements to establish how they attached to their targets. That is a meaningful research accomplishment, although it is far removed from showing that the designs are safe or effective medicines for patients.
The important point is the direction of travel. Scientists are beginning to specify a task, propose molecular designs, build physical versions and measure what happens. When those results are fed into the next round of design, scientific work can become a repeated cycle of prediction, experiment and improvement.
Why this could reshape the search for treatments
Drug development is difficult for reasons that computers alone cannot solve. A molecule may attach to the desired protein but fail to reach the correct part of the body. It might break down too quickly, affect other processes or cause side effects. A successful laboratory experiment is a beginning, not a shortcut around clinical trials and medical oversight.
Even so, the earliest research stages are enormously important. If scientists can reject weak ideas sooner and select promising molecules more intelligently, they may be able to use their time and equipment much more effectively. Automated laboratory systems could help test more of those ideas, while human researchers concentrate on interpreting unexpected results and designing the next experiment.
For diseases with few treatment options, that could eventually enlarge the range of approaches researchers can afford to explore. It would not guarantee a breakthrough for every condition, but it could change which scientific questions become practical to investigate.
THE IMPACT / IT GETS PERSONAL
What could this mean
for my future?
More possible answers to difficult diagnoses
For patients and families, the hope is not an instant personalized cure. It is a future in which researchers have more ways to investigate the exact biological process behind an illness and a better chance of finding promising treatment candidates. Some rare conditions may eventually attract research that was previously too costly or technically difficult. Whether that translates into affordable treatment will still depend on trials, approvals, manufacturing and access.
New scientific teams, new kinds of expertise
The people developing tomorrow’s treatments may include molecular biologists, physicians, computer scientists, robotics engineers and specialists who make laboratory information trustworthy. A student interested in medicine may find that learning how data and computation work becomes almost as important as understanding traditional laboratory practice. The opportunity is not to replace medical judgment but to give experts better tools for asking and answering biological questions.
Competition may shift from searching to proving
A biotechnology company might spend less of its early budget blindly testing possible molecules and more effort identifying useful targets, proving that designs work and developing reliable manufacturing processes. Large pharmaceutical companies and smaller research firms could gain access to overlapping design capabilities, making experimental quality and clinical execution more important differentiators. A commercially valuable treatment still requires years of disciplined work after an interesting computer-generated candidate appears.
The research process itself could be redesigned
Pharmaceutical research has long operated through expensive sequences of discovery, testing, refinement and clinical development. Artificial intelligence and laboratory automation could shorten some early feedback loops and enable researchers to investigate more ideas. At the same time, regulators, insurers and health systems will have to evaluate a growing range of technically novel treatments without confusing speed of discovery with evidence of safety and benefit.
From personalized medicine to engineered possibility
Jim Carroll has explored the acceleration of medical science and personalized medicine for decades. In a Genentech keynote for hundreds of executives, he examined how faster genetic knowledge and targeted therapies were beginning to change the pharmaceutical industry's assumptions. His broader point was that when the pace of scientific knowledge changes, the strategy of organizations working with that knowledge must change too.
For a biotech research leader, the immediate question is concrete: which expensive decision in the earliest phase of drug discovery could become better informed through computational design? Choose one target, define the evidence needed to judge candidate quality and compare the tool against the existing process. The objective is to learn where it genuinely improves research—not to assume that an algorithm has already invented the next medicine.
Just imagine what
becomes possible.
The WOW is not that a computer can create a cure on command. It is that we are beginning to change biological research from finding useful molecules by chance toward deliberately designing and testing some of the tools we need. That could alter the possibilities facing tomorrow’s doctors, researchers and patients.
What's real—and what's still a possibility?
The AlphaFold count refers to predicted structures. The antibody experiments demonstrate laboratory binding, not proven treatments for patients.
Read the evidence and original sources
Structure predictions, not experimentally verified treatment candidates.
Laboratory-tested binding to specified protein targets; no approved therapy established by this experiment.
How YottaBit treats evidence and uncertainty ↗
Original research references: C-02 · convergence:2
KEEP EXPLORING