An AI has drawn the first single map of the protein world
What happened
An international research team has built an artificial intelligence model that puts two very different kinds of information about proteins onto a single map. The work was published in the Proceedings of the National Academy of Sciences on 11 September 2026. The model is called CLSS, short for Contrastive Learning Sequence-Structure. The team was led by Professor Rachel Kolodny and PhD candidate Guy Yanai at the University of Haifa, Professor Nir Ben-Tal and graduate student Gabriel Axel at Tel Aviv University, and Specially Appointed Associate Professor Liam M. Longo at the Earth-Life Science Institute, part of the Institute of Science Tokyo. Proteins can be described in two ways. One is the sequence: the order of amino acids strung together, a bit like letters in a very long word. The other is the structure: the three-dimensional shape that string folds itself into. Until now, computer models have tended to handle one or the other. CLSS handles both at once, so that the sequence version and the shape version of the same protein end up in the same place on the map.
Why this matters
Proteins do almost all the practical work in a living body. They digest your food, carry oxygen in your blood, fight infection and hold your tissues together. Most medicines work by interacting with one. Two proteins can have almost nothing in common in their sequence and still fold into nearly the same shape and do nearly the same job. The opposite happens too. That is why looking at only one kind of information can mislead you. Professor Longo put it simply: "This gives us a way to look at the protein universe through sequence and structure at the same time, rather than treating them as separate worlds." There is a bigger prize behind this. Proteins are the oldest machinery on Earth. Being able to see relationships between them across the whole protein world is a route to understanding how life's building blocks evolved over billions of years, and potentially to designing new ones.
What the evidence actually says
CLSS works by turning each protein into what researchers call an embedding. That is a list of numbers that acts like a location. Similar proteins sit close together, different ones sit far apart. The clever part is the training method, contrastive learning. The model is repeatedly shown the sequence and the structure of the same protein and taught to place them at the same address. Do that across a huge number of proteins and you end up with one shared map instead of two separate ones. It is worth being clear about what this is and is not. This is a tool for seeing patterns and relationships. It is not a drug discovery machine, and it does not predict whether a medicine will work in a person. It also sits alongside, rather than replacing, structure prediction tools that work out what shape a protein will fold into. Those answer the question "what does this one look like?" CLSS answers a different question: "how does this one relate to all the others?" As with any model, the map is only as good as the data it was trained on. Parts of the protein world are much better studied than others.
Practical advice
This is basic science rather than something that changes your day. But it does offer a useful way to read AI-in-science stories, which are arriving constantly. Look for what the tool actually produces. A map of relationships, a prediction of a shape and a recommendation about a patient are three completely different things, even when all three are called AI. Look for the journal and the team. Peer-reviewed work in an established journal, from named researchers at named universities, is a different proposition from a company announcement. Be patient with timelines. Tools like this one usually make their difference years later, indirectly, by helping other researchers ask better questions. And if you are a student wondering what to study, note who built this: chemists, biologists, physicists and computer scientists working together. That mix is where a lot of modern science now happens.
What to know
Researchers from the University of Haifa, Tel Aviv University and the Institute of Science Tokyo have built an AI model called CLSS that maps proteins using their sequence and their 3D shape at the same time. It was published in PNAS on 11 September 2026. The aim is to see evolutionary relationships across the whole protein world that are hard to spot when sequence and structure are studied separately. This is early-stage, fundamental science. It is a research tool, not a treatment, and it will take years for any benefit to reach patients. Sources: AZoLifeSciences, "New AI Model Unites Protein Sequences and Structures for Evolutionary Insights", 11 September 2026, https://www.azolifesciences.com/news/20260911/New-AI-Model-Unites-Protein-Sequences-and-Structures-for-Evolutionary-Insights.aspx | Proceedings of the National Academy of Sciences, "Contrastive learning unites sequence and structure in a global representation of protein space", 11 September 2026, https://doi.org/10.1073/pnas.2532702123 | Mirage News, "Mapping Protein Universe: Sequence Meets Structure", 11 September 2026, https://www.miragenews.com/mapping-protein-universe-sequence-meets-1742636/ This article is for general information. It covers scientific research and is not health advice, and it does not replace advice from a doctor, pharmacist or other qualified healthcare professional.
Related articles
A pain-relief patch the size of a plaster, controlled from a phone
A research team in South Korea has built a small wearable device that delivers electrical pain relief through the skin, and can be controlled remotely from a sm…
Something dark is hiding in Venus's clouds, and scientists just narrowed down what it can be
Venus has a mystery that has sat unsolved for roughly a century. Seen in ordinary light, the planet is a smooth, featureless cream-yellow ball. Look at it in ul…
Scientists looked a quarter of a proton deep inside an atom - and found something odd
Physicists working on the ALICE experiment at CERN's Large Hadron Collider have managed to look deeper inside atomic nuclei than anyone has before. What they fo…