Research themes
Research
Our shared core is probabilistic and network modelling: inferring structure from noisy data, quantifying uncertainty, and reading mechanism from topology. We apply it across biomedicine, global health and beyond.
Bayesian & Network Inference
Probabilistic graphical models, MCMC samplers and causal network learning for noisy, high-dimensional data.
Our methodological core. We design Markov chain Monte Carlo samplers that escape local optima when learning Bayesian-network structure (the Neighbourhood and Hit-and-Run samplers, released in BNMCMC), fit exponential random graph models such as the p1-model to quantify how likely each interaction is in a condition-specific network, and build structured elicitation techniques (BARD) for reasoning under uncertainty.
Questions we ask
- How can we learn causal structure from small-sample omics data without getting trapped in local optima?
- Which network motifs (for example, V-structures) reveal a switch from targeted to bypass signalling?
Methods from this theme
Most-cited papers in this theme
-
BARD: A structured technique for group elicitation of Bayesian networks to support analytic reasoning
Risk Analysis 42 (6), 1155-1178 · 2022
-
Prediction of signaling cross-talks contributing to acquired drug resistance in breast cancer cells by Bayesian statistical modeling
BMC Systems Biology 9 (1), 2 · 2015
-
Bayesian model of signal rewiring reveals mechanisms of gene dysregulation in acquired drug resistance in breast cancer
PLOS ONE 12 (3), e0173331 · 2017
-
Cross-talk categorisations in data-driven models of signalling networks: a system-level view
In: Gene Expression and Regulation in Mammalian Cells - Transcription From General Aspects, IntechOpen · 2018
Cancer Systems Biology & Drug Resistance
Modelling how signalling pathways rewire and cross-talk when tumours escape targeted therapy.
Tumours that initially respond to targeted inhibitors such as lapatinib or dasatinib often relapse because signalling is reprogrammed. We model this rewiring with data-driven cross-talk networks, define formal categories of pathway cross-talk (Type-I/Type-II, XDaMoSiN), and prioritise perturbation drivers and prognostic biomarkers in breast, pancreatic, liver and lung cancers, validated against TCGA, CPTAC and proteomic cohorts.
Questions we ask
- Which compensatory pathways cross-talk with the drug target as resistance is acquired?
- Can the topology of a perturbation network nominate the next therapeutic target?
Methods from this theme
Most-cited papers in this theme
-
Prediction of signaling cross-talks contributing to acquired drug resistance in breast cancer cells by Bayesian statistical modeling
BMC Systems Biology 9 (1), 2 · 2015
-
Bayesian model of signal rewiring reveals mechanisms of gene dysregulation in acquired drug resistance in breast cancer
PLOS ONE 12 (3), e0173331 · 2017
Drug Repositioning & Network Medicine
Multi-modal drug similarity, interactome proximity and combination-therapy prioritisation.
We integrate chemical, target, functional and pathway evidence into multi-modal drug–drug similarity networks (DrugSimDB, more than 238,000 significant drug pairs across 10,000+ small molecules), and use network-proximity and separation measures on the human interactome to prioritise synergistic drug combinations, as in the COVID-CDR platform and the cyclosporine–selinexor combination for SARS-CoV-2.
Questions we ask
- Which drug pairs act on complementary, non-overlapping neighbourhoods of a disease module?
- How can heterogeneous drug evidence be fused into a single, extensible similarity network?
Methods from this theme
Most-cited papers in this theme
-
Identification of genetic biomarkers, drug targets and agents for respiratory diseases utilising integrated bioinformatics approaches
Scientific Reports 13 (1), 19072 · 2023
-
Exploring gene regulatory interaction networks and predicting therapeutic molecules for hypopharyngeal cancer and EGFR-mutated lung adenocarcinoma
FEBS Open Bio 14 (7), 1166-1191 · 2024
-
Single nucleus RNA sequencing profile analysis to reveal cell type specific common molecular drivers of Parkinson's disease and therapeutic agents
Scientific Reports 15 (1), 27086 · 2025
Multi-omics & Single-cell Bioinformatics
Biomarker discovery from bulk, single-cell and spatial omics with statistical and deep learning.
From promoter prediction and copy-number–expression integration to single-cell and single-nucleus RNA-seq, proteomics (DIA-MS) and dual RNA-seq of host–pathogen systems, we develop pipelines that turn high-throughput molecular data into interpretable biomarkers, regulatory footprints and candidate therapeutics in cancer, neurodevelopment and Parkinson's disease.
Questions we ask
- Which cell-type-specific drivers are shared across neurodegenerative conditions?
- How do host and pathogen transcriptional programmes co-evolve over the course of infection?
Methods from this theme
Most-cited papers in this theme
-
Prediction of plant promoters based on hexamers and random triplet pair analysis
Algorithms for Molecular Biology 6 (1), 19 · 2011
-
Identification of genetic biomarkers, drug targets and agents for respiratory diseases utilising integrated bioinformatics approaches
Scientific Reports 13 (1), 19072 · 2023
-
Bioinformatics and system biology approaches to identify the diseasome and comorbidities complexities of SARS-CoV-2 infection with the digestive tract disorders
Briefings in Bioinformatics 22 (6), bbab126 · 2021
Trustworthy Medical AI
Explainable, federated and efficient deep learning for imaging, physiological signals and clinical data.
We build AI systems that clinicians can trust and deploy: explainable models (SHAP, LIME, attention) for chronic kidney disease, thyroid, diabetes and heart failure; federated and graph-federated learning for privacy-preserving ECG diagnosis; and transformer and CNN architectures for MRI, CT, X-ray, ultrasound, EEG and wearable-sensor data, often in support of people living with disability.
Questions we ask
- How do we keep accuracy while making models auditable, private and lightweight enough for the point of care?
Most-cited papers in this theme
-
Deep CNN-LSTM with self-attention model for human activity recognition using wearable sensor
IEEE Journal of Translational Engineering in Health and Medicine 10, 1-16 · 2022
-
A classification of MRI brain tumor based on two stage feature level ensemble of deep CNN models
Computers in Biology and Medicine 146, 105539 · 2022
-
Emotion Recognition from EEG Signal Focusing on Deep Learning and Shallow Learning Techniques
IEEE Access 9, 94601-94624 · 2021
-
LungNet: A hybrid deep-CNN model for lung cancer diagnosis using CT and wearable sensor-based medical IoT data
Computers in Biology and Medicine 139, 104961 · 2021
Global Health & Epidemiology
Population-scale burden estimation and real-time epidemic analytics.
As a Senior Collaborator of the Global Burden of Disease (GBD) study at the Institute for Health Metrics and Evaluation (IHME), we contribute to the analyses that underpin global health policy, from cancers and cardiovascular disease to infections and maternal health. We also build real-time forecasting tools (COVIDSpread) and machine-learning analyses of COVID-19 severity, mortality and vaccine reactogenicity.
Questions we ask
- How can forecasting and burden estimates be made transparent and actionable for policy makers?
Methods from this theme
Most-cited papers in this theme
-
Global, regional, and national burden of breast cancer among females, 1990-2023, with forecasts to 2050: a systematic analysis for the Global Burden of Disease Study 2023
The Lancet Oncology 27 (3), 302-326 · 2026
-
Machine Learning Approach to Predicting COVID-19 Disease Severity Based on Clinical Blood Test Data: Statistical Analysis and Model Development
JMIR Medical Informatics 9 (4), e25884 · 2021
-
Global burden of lower respiratory infections and aetiologies, 1990-2023: a systematic analysis for the Global Burden of Disease Study 2023
The Lancet Infectious Diseases 26 (4), 343-361 · 2026
-
Machine Learning Approaches to Identify Patient Comorbidities and Symptoms That Increased Risk of Mortality in COVID-19
Diagnostics 11 (8), 1383 · 2021
AI Foundations & Applied AI
New optimisers and loss functions, vision-language agents and physics-informed learning.
Methods that generalise beyond biomedicine: the SPARK optimiser and the AGL loss for more stable training; VLM-Nav for mapless UAV navigation with vision-language models; physics-informed surrogates for polymer composites and concrete; and applied systems for agriculture (PlantCareNet), land-value analytics with differential privacy (LandSin), astronomy, cyber-security and smart-city sensing.
Questions we ask
- Can feedback from the loss trajectory make optimisation robust to the choice of learning rate?
Methods from this theme
Most-cited papers in this theme
-
SafetyMed: A Novel IoMT Intrusion Detection System Using CNN-LSTM Hybridization
Electronics 12 (17), 3541 · 2023
-
BARD: A structured technique for group elicitation of Bayesian networks to support analytic reasoning
Risk Analysis 42 (6), 1155-1178 · 2022
-
A Human-Robot Interaction System Calculating Visual Focus of Human's Attention Level
IEEE Access 9 · 2021
-
PlantCareNet: an advanced system to recognize plant diseases with dual-mode recommendations for prevention
Plant Methods 21 (1), 52 · 2025