A collaborative team of computational biologists, data designers, and software engineers at Blackbridge Institute Polytechnic and Arts has announced the open-source release of ‘Proteus Atlas,’ a next-generation visualisation and analysis platform for exploring complex protein-protein interaction (PPI) networks. Detailed in a feature article in a prestigious journal of computational biology, the platform addresses a fundamental challenge in modern drug discovery: the difficulty of predicting how a new therapeutic molecule might cause unintended side effects. Proteus Atlas combines a powerful graph-based analytical engine with a novel, intuitive interface designed to make the dizzying complexity of the cellular world more navigable for researchers.
The human body functions through a vast, intricate network of proteins interacting with one another in a dynamic, ever-shifting dance. This ‘interactome’ is the landscape where diseases manifest and where medicines must intervene. However, a drug designed to target one specific protein often interacts with dozens of others—so-called ‘off-target’ effects—leading to unforeseen side effects. Current tools for mapping these interactions are often static, cumbersome, and fail to represent the multi-dimensional nature of these biological systems, a limitation the Proteus Atlas project was created to overcome.
The platform’s foundation is a sophisticated backend built by a postgraduate research group within the Computational Engineering & Intelligent Systems discipline. It utilises graph neural networks (GNNs), a specialised form of artificial intelligence adept at learning from network-structured data. Unlike older methods that treat the interactome as a fixed map, the GNN-powered engine models it as a probabilistic system, capable of predicting how interactions might shift or change under different conditions, such as the introduction of a new drug compound.
“We are moving from a static photograph to a dynamic weather forecast,” explains a newly appointed Professor of Computational Biology who leads the biological side of the project. “The cell is not a fixed circuit board; it’s a crowded, chaotic, and constantly reorganising party. Our model learns the underlying rules of this party—which proteins are likely to interact, under what conditions, and how strongly. This allows us to run simulations that were previously computationally prohibitive.”
The most striking innovation, however, is the platform’s user interface, a result of a deep collaboration with the Digital Media & Communication Design department. Recognising that even the most powerful data is useless if it cannot be understood, the team spent eighteen months developing a novel visualisation paradigm. Instead of presenting the network as a dense, unreadable ‘hairball’ of nodes and lines, Proteus Atlas employs a multi-layered, three-dimensional interface. Researchers can zoom through the network, filter interactions based on confidence scores or cellular location, and cluster protein families into visually distinct, explorable communities.
A Master’s student in Digital Communication Design who worked on the user experience (UX) described the central design challenge: “Our goal was to manage cognitive overload. The human brain isn’t equipped to process ten thousand connections at once. We used a combination of force-directed layout algorithms and hierarchical clustering to create a natural sense of order. The interface actively simplifies the view at a distance but reveals layers of rich detail as you dive deeper. We wanted it to feel less like reading a chart and more like navigating a celestial map.”
Furthermore, the platform integrates a unique predictive module designed to flag potential safety concerns early in the discovery process. By analysing the topological properties of a proposed drug target within the wider network, the system identifies proteins that are ‘topologically similar’—meaning they occupy similar positions in the network and are therefore more likely to be affected by the same drug. This serves as an early-warning system for potential off-target effects.
The team is careful to manage expectations about this feature. “It is absolutely not a deterministic oracle,” cautions the lead engineering researcher. “It’s a probabilistic spotlight. It can’t tell you a drug will cause a specific side effect. But it can tell you, with a quantifiable degree of confidence, ‘This target shares network characteristics with twelve other proteins involved in crucial cardiac functions; you should prioritise investigating this in the lab.’ It’s about focusing limited experimental resources where they are most needed, a principle of harm reduction applied at the molecular level.”
The Proteus Atlas is being released to the academic community as an open-source beta platform. The team acknowledges that the immense computational resources required to run the full GNN model is a current limitation, but they are already working on a lighter version for broader accessibility. By making the code and their foundational research public, they aim to crowdsource improvements and foster a new, more transparent standard for computational drug discovery. The project is a profound statement on the Institute’s belief that the greatest scientific breakthroughs of this century will occur not within isolated disciplines, but in the challenging, dynamic, and ultimately fruitful spaces between them.
Leave a Reply