Research

Our research is cross-disciplinary, combining neuroscience with mathematics, engineering, computer science and artificial intelligence. We develop computational models to understand the principles underlying brain organisation and function, and translate these insights into new approaches for brain and mental health. Our lab research focuses on:

• NeuroAI and brain modelling to understand how brain networks interact to support cognition, and to develop generative and AI models of brain organisation and dynamics;
• Biological and neuroscience-inspired intelligence to uncover computational principles underlying learning, reasoning, planning and adaptive behaviour;
• Precision neurotherapeutics, including psychedelics and brain stimulation, combining brain imaging, computational modelling and AI to understand mechanisms and predict individual treatment response.

Areas of research/projects:

Mechanistic computational neuroscience

We develop advanced computational and theoretical tools to understand how the brain is organised. Our research is geared towards developing new dynamic causal models (DCM) that can explain how the brain’s measured data is caused. We use these models to integrate multi-modal empirical measurements from functional, structural and diffusion magnetic resonance imaging (MRI).  In this stream of work we are:

• Developing multi-scale models of brain function
• Integrating multimodal brain imaging using generative models
• Identifying mechanisms underlying brain network organisation and dysfunction

_{Three complementary approaches for integrating structural connectivity into models of directed effective brain interactions: Bayesian priors, embedded structural constraints, and machine-learning-based estimation. Figure from Greaves et al., Nat. Rev. Neurosci. 2025}

NeuroAI

We develop artificial intelligence approaches for modelling complex brain dynamics across individuals, imaging modalities and experimental conditions. Our work combines neuroscience, machine learning and generative modelling to build scalable models that can learn representations of brain organisation from large and heterogeneous datasets. This includes multimodal foundation models such as BrainSymphony and deep generative models such as TAVRNN. Our research aims to:

• Develop multimodal foundation models for brain imaging
• Learn generalisable representations of brain dynamics across datasets and modalities
• Use AI to predict brain states, individual differences and clinically relevant outcomes

_{Architecture of BrainSymphony, a multimodal foundation model that integrates spatial, temporal and local features from fMRI with structural connectome information to learn compact brain representations for downstream prediction.}

Biological intelligence

We investigate how biological neural systems learn, adapt and behave intelligently, and how these principles can inform our understanding of both brains and artificial intelligence. Our research spans active inference, generative models and synthetic biological intelligence, including neuronal cultures interacting with environments through closed-loop systems. By studying intelligence across different biological scales and substrates, we aim to identify general principles underlying adaptive behaviour. In this stream of work we are:

• Studying learning and adaptive behaviour in biological neural systems
• Developing closed-loop models of synthetic biological intelligence
• Using active inference to understand principles of learning, reasoning and planning

_{DishBrain integrates human or rodent cortical neurons with a simulated Pong environment through a high-density multielectrode array, enabling closed-loop learning and adaptive behaviour as a model of synthetic biological intelligence. This Figure is taken from Kagan et al., Neuron, 2022}

Precision neurotherapeutics

We combine neuroimaging, computational modelling and clinical trials to understand and improve emerging treatments for brain and mental health conditions. Our work includes psychedelic-assisted therapies and brain stimulation, with a particular focus on identifying mechanisms of treatment response and explaining individual variability. Ultimately, we aim to move from understanding average treatment effects towards predicting which intervention is most likely to benefit an individual. Our research asks:

• How do neurotherapeutic interventions alter brain dynamics?
• Which neural mechanisms predict treatment response and durable recovery?
• Can brain imaging and AI enable personalised treatment selection and optimisation?

_{Context-dependent brain activity under psilocybin forms distinct neural trajectories that become more strongly aligned with ongoing experience and scale with the intensity of subjective psychedelic effects.}