Capturing domain-level detail of molecular interactions in the IMEx Consortium binding domain data set
This paper introduces the IMEx binding domain dataset, a large-scale, manually curated collection of experimentally derived protein interaction regions mapped to UniProtKB sequences, which serves as a gold standard for benchmarking interface prediction methods and elucidating the mechanistic basis of disease-associated mutations.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Inside every living cell, a vast network of conversations takes place constantly. Proteins, the workhorses of life, do not act alone; they must physically touch and connect with other molecules to send signals, build structures, and carry out the chemical reactions that keep an organism alive. These connections are not random collisions but highly specific meetings, often occurring at precise spots on the protein's surface. Just as a key fits into a specific lock, a protein uses a particular region of its shape to grab onto a partner. Scientists call these specific grabbing spots "binding regions." Understanding exactly where these spots are located is crucial because it reveals how cells function normally and, just as importantly, how they break down when disease strikes. If a genetic mutation changes the shape of a binding region, the protein might lose its ability to connect, causing the entire cellular conversation to fail.
For over twenty years, a global team of scientists has been meticulously collecting evidence of these specific meeting points. This effort, known as the IMEx Consortium, involves experts who read thousands of scientific papers to extract details about how proteins interact. They do not just record that two proteins touch; they identify the exact sequence of building blocks, known as amino acids, that make the connection possible. In a recent update to this massive collection, the team has released a comprehensive dataset containing nearly 233,000 of these experimentally verified binding regions. This new resource acts as a detailed map, linking specific parts of a protein's sequence to the interactions they mediate, and connects this information to the broader landscape of protein families and human genetic diseases.
The researchers behind this dataset, working across Europe and beyond, have spent decades curating this information from peer-reviewed literature. They organized the data into three main categories based on how the experiments were conducted. The largest group, containing over 179,000 entries, represents "sufficient binding regions." These are segments of a protein that, when isolated and tested, are enough to cause a connection with a partner. A second group, numbering over 48,000, describes "binding-associated regions," which are areas suspected to be involved in the interaction but where the exact boundaries are less clear. The smallest group, with about 5,300 entries, lists "necessary binding regions." These are the most precise annotations, identifying the specific amino acids that are absolutely required for the connection to happen. The data covers interactions in more than 1,000 different species, though the majority of the information comes from well-studied organisms like humans, mice, fruit flies, and yeast.
To make this data more useful for modern research, the team developed a new method to automatically link these binding regions to known protein families. While human curators had previously added these links only when a paper explicitly named a specific domain, the new automated pipeline scans the entire dataset to find overlaps. The researchers found that nearly half of the human binding regions contain at least one known protein domain, but in many cases, the binding region extends well beyond the boundaries of that domain. This suggests that the "grip" a protein uses to hold onto a partner is often larger and more complex than the standard modular parts scientists usually study. By systematically matching these regions to a database of known protein structures, the team was able to assign a domain identity to thousands of binding regions that had previously lacked one, effectively filling in gaps in our understanding of protein architecture.
The team also explored how these binding regions relate to short, simple patterns in protein sequences called linear motifs. Unlike large, complex domains, these motifs are tiny sequences of just a few amino acids that can trigger interactions. The analysis revealed that many of the binding regions in the dataset correspond to these known motifs, particularly those involved in docking sites or chemical modifications like phosphorylation. This finding bridges the gap between large structural domains and the tiny, fleeting interactions that often regulate cell behavior. It confirms that the dataset captures a wide spectrum of interaction types, from the sturdy, well-defined domains to the flexible, short sequences that allow for rapid cellular responses.
Perhaps the most significant application of this dataset lies in understanding human disease. The researchers mapped thousands of genetic variants, including those known to cause disease, onto these binding regions. They found that nearly 12,000 disease-associated mutations fall directly within experimentally defined binding regions. This is a powerful insight because it moves beyond simply knowing that a mutation is harmful; it explains why. By pinpointing the mutation to a specific interaction interface, scientists can see that the disease likely arises because the protein can no longer connect with its partner. The study highlighted that these disruptive mutations are common in conditions ranging from immune disorders and hormonal imbalances to various forms of cancer and neurodegenerative diseases. For variants whose effects are currently unknown, placing them within a binding region provides a concrete starting point for further investigation, helping researchers prioritize which mutations to study first.
The dataset is freely available to the scientific community, updated with every new release of the IntAct database, and can be accessed in formats designed for both human reading and computer analysis. It serves as a gold standard for testing new tools that predict how proteins interact or how genetic changes might affect health. While the data reflects the biases of the scientific literature, with a heavy focus on well-studied proteins and a few large-scale studies, it remains the most complete collection of its kind. By providing a clear, experimentally verified map of where proteins connect, this work offers a foundational resource for deciphering the molecular logic of life and the specific ways that logic can go wrong in disease.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.