Back to news

AI for Tomorrow's Breakthroughs: 15 New Projects Funded to Transform Research

27 November 2025

The Accelerate Programme for Scientific Discovery and the Cambridge Centre for Data-Driven Discovery (C2D3) are proud to announce funding for 15 groundbreaking projects that will use artificial intelligence to tackle some of the most pressing challenges in science and society.

From detecting Alzheimer’s disease years before symptoms emerge, to identifying coral reefs that can survive climate change, to revolutionising plastic recycling—these interdisciplinary teams will develop AI tools with the potential to deliver transformative benefits across health, environment, and society.

Detecting Disease Before Symptoms Appear

Several projects are pushing the boundaries of early disease detection, using AI to spot warning signs long before traditional methods.

A Brain Imaging Foundation Model For Clinical Diagnosis in Dementia will train AI on massive brain imaging datasets to detect Alzheimer’s, vascular dementia, and other subtypes years before symptoms appear. With over 57 million people affected by dementia worldwide, earlier detection could enable intervention when treatments are most effective.

In a related project exploring the relationship between sex hormones and brain changes in people at increased genetic risk of dementia, researchers will use machine learning to analyse a decade of brain scans, genetic data, and hormone measurements to identify which women at higher dementia risk could benefit from targeted hormone replacement therapy.

Coronary artery CT foundation model (ViTAL-CT) tackles the world’s leading cause of death by training AI on 50,000 heart scans to automatically detect high-risk blockages and predict future heart attacks, delivering faster and more consistent diagnoses than current manual analysis.

Accelerating Cancer Research and Treatment

Multiple projects harness AI to understand and combat cancer from novel angles.

The AI-Powered Virtual Gut project will develop the first digital twin of the colon—a “Virtual Gut” that simulates how colon cancer develops over time by training neural networks on millions of individual cell measurements. This could reveal exactly when and how healthy cells become cancerous, identifying critical windows for early intervention.

An innovative AI in Cancer Genomics Hackathon will train early-career researchers to use machine learning to study cancer-resistant species like whales, elephants, and naked mole-rats, uncovering natural tumour-fighting mechanisms that could inspire new approaches to preventing and treating human cancers.

Researchers working to unlock mechanistic insight into the drivers of dementia will develop an AI framework using convolutional neural networks to analyse microscope images of brain tissue and instantly calculate disease progression rates—packaged as a free tool to accelerate the search for effective treatments.

Understanding Climate and Environmental Challenges

AI is also being deployed to address urgent environmental questions.

The spatial organisation of coral-algae symbiosis project uses AI with graph-based methods to analyse how algae are organised within coral cells. These spatial patterns shift before bleaching occurs, offering early warning signs and helping identify which coral-algae partnerships can survive climate change.

To help communities adapt to rising temperatures, researchers are exploring effective community-level cooling strategies in response to heatwaves using AI-powered simulations that mimic how people respond to heat. This will enable testing of different cooling approaches, such as like optimal placement of cooling centres, before implementing them in real communities.

The AI for healthy and sustainable cities project will analyse street camera images and travel patterns across 45 cities worldwide, automatically detecting cyclists and pedestrians to create detailed neighbourhood maps that help city planners promote healthier, more sustainable transportation.

Advancing Sustainable Innovation

AI-Guided Precision Solvent Selection for Sustainable Plastic Recycling addresses a major recycling challenge: different plastics require different solvents, and finding the right match wastes time and resources. Machine learning trained on over 30,000 plastic-solvent combinations will predict which solvents selectively dissolve specific plastics from waste mixtures, enabling efficient recycling that supports a circular economy.

Building Research Capacity and Tools

Several projects focus on making AI more accessible and trustworthy for researchers.

Bridging the Skills Gap: Computer Vision Training for Marine Ecologists will run a two-day workshop teaching marine ecologists with little programming experience how to use AI tools that automatically identify species and map habitats from underwater images, addressing a major bottleneck in marine ecology research.

Trusting Qualitative AI: The Cambridge Protocol develops the first quality assurance framework for AI-assisted qualitative research. Tested on interviews with people who have experienced psychosis, it will ensure AI outputs meet rigorous standards for accuracy, inclusivity, and trustworthiness in healthcare and social science.

CogMap: An Educational AI Tool for Interpreting Brain-Based Cognitive Profiles will make cutting-edge neuroscience accessible to students, researchers, and medical trainees by using machine learning to compare individual brain scans against population norms and employing language models to explain findings in simple terms.

Revealing Hidden Biological Mechanisms

Other projects are using AI to uncover fundamental biological processes.

Development of an AI-Based Platform for Quantifying Feeding Behaviour in Mice will use AI-powered video analysis to track mouse behaviour while recording brain activity from neurons controlling reproduction, feeding, and sleep, showing how these systems interact to regulate fertility and potentially identifying new treatment approaches for women experiencing fertility problems linked to poor sleep or irregular eating.

Plant automatic cell lineage reconstruction will train a neural network to recognise “sister cells” in mature plant leaves and reconstruct their complete growth history from a single image, eliminating the need for damaging repeated imaging and enabling scientists to study how temperature and other factors influence plant development.

About the Funding Programme

This year’s call received over 100 applications from teams across the University of Cambridge. Successfully deploying AI to tackle real-world challenges requires effective interdisciplinary collaboration, supported by time and resources to bring together research partners, develop new tools, and build skills and networks.

The Accelerate-C2D3 funding programme helps fill a gap in routine funding by offering small grants that can be deployed flexibly to start or scale interdisciplinary projects advancing the use of data science and AI for research and innovation.

Projects will be delivered over the next year, with updates shared through the Accelerate Programme blog. Accelerate Science and C2D3 will provide technical support and facilitation as part of project delivery.

Translating Research to Real-World Impact

To help translate promising approaches into real-world impact, Accelerate and C2D3 have for the second time offered Impact funding. Six projects from previous funding rounds are receiving support to scale their work:

• Dr Boris Bolliet (Department of Physics) is developing his model for automating data analysis with LLM agents into a robust platform providing users with personalised research environments integrated with AI agents, in collaboration with partners at Google DeepMind and the Flatiron Institute.
• Dr Anna Breger (Department of Applied Mathematics and Theoretical Physics) is developing an open software toolbox to allow wider uptake of machine learning algorithms for reconstructing degraded historical sheet music.
• Charles Emogor and Anil Madhavapeddy (Departments of Zoology and Computer Science and Technology) are working with rangers in protected areas to develop and validate AI models that predict illegal hunting activity and optimise anti-poaching patrol routes.
• Megan Ennion and Steve Watson (Faculty of Education) are delivering events to create opportunities for dialogue, knowledge exchange, and interdisciplinary collaboration on AI in Education.
• Maya Juman and Olivier Restif (Department of Veterinary Medicine) are applying their AI framework for studying RNA viruses in fruit bats to coronaviruses, contributing towards pharmaceutical innovation.
• Dr Mairead Healy (Department of Psychology) is advancing the diagnosis and treatment of OCD by validating a machine learning model with real UK clinical populations and launching an accessible clinical tool.

Projects awarded in 2025:

AI-Guided Precision Solvent Selection for Sustainable Plastic Recycling

Zheng Jie Liew, Marie Curie Research Fellow, Department of Chemical Engineering and Biotechnology

Alexei Lapkin, Professor of Sustainable Reaction Engineering, Department of Chemical Engineering and Biotechnology

Recycling mixed plastics is difficult because different types of plastic require different solvents and finding the right match through trial-and-error wastes time and resources. This project will make use of a machine learning framework trained on over 30,000 plastic-solvent combinations to predict which solvents will selectively dissolve specific plastics from waste mixtures, enabling efficient recycling without destroying material value and supporting a circular economy for plastics.

The relationship between sex hormones and brain changes in people at increased genetic risk of dementia

Axel Laurell, Clinical Research Associate, Department of Psychiatry

Maria-Eleni Dounavi, Senior Postdoctoral Researcher, Department of Psychiatry

Early menopause may increase dementia risk, possibly through hormonal changes affecting the brain, but it’s unclear which women might benefit from hormone replacement therapy. This project uses machine learning to analyse 10 years of brain scans, genetic data, and hormone measurements from 100 middle-aged participants to identify patterns linking hormones to brain changes and predict which women at higher dementia risk could benefit from targeted treatment.

Bridging the Skills Gap: Computer Vision Training for Marine Ecologists

Emily Mitchell, Assistant Professor, Department of Zoology

Underwater imaging has generated vast datasets essential for understanding marine ecosystems, but analysing this data manually remains a bottleneck in marine ecology research. The project team will run a two-day workshop teaching marine ecologists with little to no programming experience how to use computer vision — AI tools that automatically identify species and map habitats from images — so researchers can study ocean ecosystems faster and more efficiently using freely available tools.

AI in Cancer Genomics Hackathon: What can we learn from cancer-resistant species to better understand and tackle human cancer susceptibility

Laura Machesky, Professor of Biochemistry, Department of Biochemistry

Eloise Trabut, Cancer Research UK Cambridge Centre Programme Manager for Fundamental Biology of Cancer, Department of Oncology

Whales, elephants, and naked mole-rats rarely get cancer despite their size and long lives. The project team will deliver a hackathon hosted by the Cancer Research UK Cambridge Centre training early-career researchers to use machine learning and comparative genomics to analyse these animals’ DNA, uncovering natural tumor-fighting mechanisms that could reveal new approaches to preventing and treating human cancers.

Exploring effective community-level cooling strategies in response to heatwaves on population health: an agent-based modelling simulation study

Yuanfei Liu, PhD Student, Department of Psychiatry

Sharon Neufeld, Senior Research Associate, Department of Psychiatry

Heatwaves are among the most lethal climate-related hazards with rising risks under climate change, but testing different cooling strategies — like where to place cooling centers — is expensive and slow. This project uses AI-powered computer simulations that mimic how people respond to heat to test which cooling approaches work best before implementing them in real communities, providing a cost-effective tool which enhances community resilience and reduces heat-related illness.

Unlocking mechanistic insight into of the drivers of dementia

Matthew Cotton, Postdoctoral Research Associate, Department of Chemistry

Professor Sir David Klenerman, Royal Society GSK Research Professor, Department of Chemistry

Biophysics-informed brain simulations can model how dementia spreads through the brain but analysing patient tissue samples requires expensive computing power and specialist skills, slowing research. This project will develop an AI-driven simulation-based inference framework using convolutional neural networks to analyse microscope images of brain tissue directly and instantly calculate disease progression rates, packaged as a free tool that any researcher can use, accelerating the search for effective treatments.

AI-Powered Virtual Gut: A Discovery Platform for Early Intervention in Colorectal Cancer

Qiuyu Lian, Senior Research Associate, Gurdon Institute & Deparment of Applied Mathematics and Theoretical Physics

Cancer takes years to develop, but most studies only capture snapshots at diagnosis, missing critical early changes when treatment could be most effective. This project aims to develop the first “Virtual Gut”— a digital twin of the colon developed by training large neural-network models on millions of individual cell measurements to simulate how colon cancer develops over time—revealing exactly when and how healthy cells become cancerous and identifying windows for early intervention.

Trusting Qualitative AI: The Cambridge Protocol

Ben Laws, Research Associate, Department of Psychiatry

Large language models can rapidly analyse interview transcripts and survey responses, but their outputs remain prone to errors, bias, and hallucination with no established quality assurance standards to ensure rigour, validity, and inclusivity in research. This project develops the first quality assurance framework for AI-assisted qualitative research, tested on interviews with people who have experienced psychosis, ensuring AI outputs meet rigorous standards for accuracy, inclusivity, and trustworthiness in healthcare and social science.

Coronary artery CT foundation model (ViTAL-CT)

Dr Yuan Huang, Research Associate, Department of applied mathematics and theoretical physics

James Rudd, Professor of Cardiovascular Medicine, Department of Medicine

Coronary artery disease is the world’s leading cause of death. Heart scans can detect dangerous artery blockages, but analysing them is slow, requires specialist training, and can suffer from different interpretation of findings by different medical professionals. This project trains an AI model specifically designed for heart arteries on 50,000 scans to automatically detect high-risk blockages and predict future heart attacks, delivering faster, more consistent diagnoses and ultimately reducing heart attack incidence.

AI for healthy and sustainable cities

Kyriaki Kokka, Research Assistant, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine

James Woodcock, Professor of Transport and Health Modelling, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine

City planners need detailed data on walking and cycling to build better infrastructure, but traditional surveys can’t provide neighbourhood-level information. This project uses an AI framework to analyse street camera images and travel patterns across 45 cities worldwide, automatically detecting cyclists and pedestrians to create detailed neighbourhood maps showing how people move, helping planners promote healthier, more sustainable transportation.

The spatial organisation of coral-algae symbiosis

Susie McLaren, Postdoctoral Researcher, Gurdon Institute

Coral reefs depend on partnerships between coral and algae, but these break down under heat stress causing bleaching. This project uses AI with graph-based methods to analyse how algae are spatially organized within coral cells, patterns that shift before bleaching occurs, revealing early warning signs and identifying which coral-algae partnerships can survive climate change to inform coral reef conservation efforts.

CogMap: An Educational AI Tool for Interpreting Brain-Based Cognitive Profiles

Richard Bethlehem, Associate Professor of Neuroinformatics, Department of Psychology

Marcella Montagnese, Junior Research Fellow in Biological and Medical Sciences, Christ’s College Cambridge & Research Associate, Department of Psychology

Brain imaging research generates valuable insights but requires specialist expertise to interpret. This project will develop CogMap, an educational tool that uses machine learning to compare individual brain scans against population norms, identify patterns linked to memory and thinking abilities, and employs a language model to simply explain findings and suggest personalised cognitive tests, making cutting-edge neuroscience accessible to students, researchers, and medical trainees.

Development of an AI-Based Platform for Quantifying Feeding Behaviour in Mice to Uncover Hypothalamic Crosstalk between Sleep, Fertility, and Energy Balance

Szilvia Vas, Postdoctoral Research Associate, Department of Physiology, Development & Neuroscience

Women with poor sleep or irregular eating often experience fertility problems, yet the underlying mechanisms linking these disturbances to reduced fertility remain largely unknown. This project will make use of AI-powered video analysis to track mouse behaviour in unprecedented detail while recording brain activity from neurons controlling reproduction, feeding, and sleep, revealing exactly how these systems interact to regulate fertility and potentially identifying new treatment approaches.

A Brain Imaging Foundation Model For Clinical Diagnosis in Dementia

Zahara Gironés Delgado-Urena, Research Associate, Department of Clinical Neuroscience

Dementia affects over 57 million people worldwide, but cases aren’t diagnosed until symptoms are severe and new treatments are less effective, yet brain changes begin years earlier. This project will develop an AI foundation model trained on massive brain imaging datasets that can detect Alzheimer’s, vascular dementia, and other subtypes of dementia years before symptoms appear by analysing brain scans, cognitive tests, and clinical information together, enabling earlier intervention when treatments work best.

Plant automatic cell lineage reconstruction

Elise Laruelle, Research Associate, Sainsbury Laboratory

Sarah Robinson, Research Group Leader, Sainsbury Laboratory

Understanding how plants grow requires tracking individual cells over weeks, demanding repeated imaging that stresses plants and limits research. This project aims to train a neural network to recognize “sister cells” in mature plant leaves and reconstruct their complete growth history from a single image, eliminating damaging repeated imaging and enabling scientists to study how temperature and other factors influence plant development.