Artificial Intelligence continues to push the boundaries of discovery — not just in data and automation, but in biology too. In a groundbreaking development, scientists at the University of Glasgow have created a new AI model that can interpret the “language” of proteins, potentially transforming how we understand diseases and virus behavior.
AI Enters the World of Biology
Artificial Intelligence is now not just reading data — it’s learning the language of life itself.
In a breakthrough that merges AI, biotechnology, and medical science, researchers from the University of Glasgow have developed a powerful new model called PLM Interact — an advanced Protein Language Model (PLM) designed to decode how proteins “talk” to one another inside cells.
This new system could revolutionize how scientists understand diseases, infections, and even virus evolution.
What Is PLM-Interact?
The PLM-Interact model, recently published in Nature Communications, uses large-scale machine learning to interpret protein-to-protein interactions (PPIs) — the molecular conversations that drive almost every biological function.
According to the research team, PLM-Interact can predict protein interactions 16% to 28% more accurately than other state-of-the-art AI tools, including Google DeepMind’s AlphaFold3.
Even more impressively, the model can forecast how mutations impact protein interactions — a discovery that could help identify disease-causing genetic changes and potential drug targets.
Powered by Supercomputing: The Role of DiRAC
To achieve this feat, the scientists harnessed Tursa, a high-performance supercomputer from the UK’s DiRAC facility, originally designed for astrophysics and quantum simulations.
This GPU-based system helped the researchers process 421,000+ protein pairs with over 650 million parameters, allowing PLM-Interact to reach unprecedented predictive accuracy.
In the words of Dr. Ke Yuan, one of the lead researchers:
“It’s fascinating to think that a supercomputer used to explore the universe has helped us understand the inner workings of life at the molecular level.”
Predicting Disease and Virus Behavior
The research goes beyond theoretical AI modeling.
PLM-Interact has shown strong performance in predicting how human and virus proteins interact, opening doors for virus emergence prediction and pandemic preparedness.
Prof. David L Robertson, from the University’s Centre for Virus Research, highlighted how valuable this model could become:
“During the COVID-19 pandemic, the urgency to understand virus-host interactions was immense. PLM-Interact could help future scientists predict these interactions faster and more accurately — enabling quicker treatment development.”
Why Protein Interactions Matter
Proteins are essential for every biological process. They are the main structural components of all cells and viruses, acting like molecular machines that interact to perform vital tasks.
When these interactions fail, diseases can emerge — from genetic disorders to cancer.
Traditionally, mapping these interactions in labs is slow and costly, but AI-driven systems like PLM-Interact can significantly accelerate discoveries by simulating millions of potential combinations in minutes.
Comparing AI Protein Models
| Model Name | Developer | Focus Area | Accuracy in Protein Interaction Prediction |
|---|---|---|---|
| PLM-Interact | University of Glasgow | Human & Viral Proteins | ✅ 16–28% higher than competing models |
| AlphaFold3 | Google DeepMind | Protein Folding | Limited PPI accuracy |
| ESMFold | Meta AI | Protein Structure Prediction | Moderate |
| ProtT5 | Rostlab | Protein Sequence Understanding | Moderate |
PLM-Interact’s ability to predict five key protein interactions — compared to AlphaFold3’s one — demonstrates its superior comprehension of molecular communication.
Broader Implications for Medicine
With its ability to analyze mutations and interactions, PLM-Interact could:
Support early disease detection
Improve drug discovery efficiency
Help design next-generation vaccines
Predict viral adaptation and pandemic potential
The model bridges the gap between AI computing and molecular biology — signaling a new era of computational medical science.
Research Team and Funding
The interdisciplinary project was led by:
Dr. Ke Yuan, School of Cancer Sciences, University of Glasgow
Prof. Craig Macdonald, School of Computing Science
Prof. David L Robertson, MRC-University of Glasgow Centre for Virus Research
The research received funding from the EU Horizon 2020, Medical Research Council, Cancer Research UK, Prostate Cancer UK, and the Biotechnology and Biological Sciences Research Council (BBSRC).
About DiRAC and Tursa Supercomputer
The work was powered by DiRAC’s Tursa — a 712-GPU cluster delivering five times more scientific throughput using half the power of its predecessor.
Tursa’s liquid-cooled GPU nodes and 5PB Lustre storage make it one of the UK’s most energy-efficient HPC systems.
About the University of Glasgow
Founded in 1451, the University of Glasgow is a member of the Russell Group and ranks 76th in the QS World University Rankings 2024.
It’s home to eight Nobel Laureates, two UK Prime Ministers, and ten Fellows of the Royal Society.
The university is deeply committed to sustainability — having pledged carbon neutrality by 2030 and divestment from fossil fuels by 2024.
For more information, visit the University of Glasgow.
Source & Disclaimer
Source: Press Release by NewsVoir
Disclaimer: This article is based on a syndicated press release from NewsVoir. TechMitra has edited and optimized the content for clarity, readability, and SEO. All facts, claims, and data are the responsibility of the source organization.
