Biomarker discovery is a challenging area that requires a multidisciplinary approach in which clinicians, laboratory-based researchers and data scientists must engage.
From experimental design to biomarker candidate list, we are able to offer support at different stages of the pipeline for biomarker discovery; from concept to candidate list:
- Study design: advice on best biological material to use, most adequate cohort, most suitable omics and help with power calculations and randomisation design
- Data processing including normalisation and batch corrections
- Exploratory analyses including uni- and multivariate statisics
- Variable selection via machine learning algorithms including random Forests (or other decision trees) or sparse partial least squares discriminant analysis
- Modelling and multi-omics integration.
We have provided our expertise in the following projects:
We are leading the computational biology aspect of the TransBioLine international consortium, which aims to explore and qualify new biomarkers for drug development and disease diagnostics.
Circulating miRNAs are potential biomarker candidates, with several advantages over other types of biomarkers. To deal with the large and complex datasets resulting from circulating miRNA profiles generated by the consortium, we are developing computational models predictive of drug-induced organ injury or disease, through the application of machine learning methods.
Additionally, to support the biological understanding of circulating miRNA signatures, a systems biology platform is being developed to enable mapping of these signatures to biological pathways and disease phenotypes. Specifically, miRNA-gene networks will be built that can be probed for their relevance to specific disease-related pathways and molecular switches to further our understanding of the mechanism by which circulating miRNAs are informative of specific processes or organ/cell injury.
Diagnostic meta-analysis for differentiation of viral and bacterial biomarkers
In this collaborative project with the DSTL (Defence Science and Technology Laboratory), our aim is to use readily available data from different platforms to find biomarkers for improving diagnosis in blood disease. We also aim to find differences in gene expression between healthy, bacterial and viral patients using artificial intelligent techniques.
Biomarker discovery for human diseases
- Identification of key Rho GTPases associated with glioma (Clarke, K. et al., PloS Genet, 2015)
- Inflammatory signals linked to the dysregulation of central metabolism gene signature of Chronic Obstructive Pulmonary Disease (COPD). (Davidsen, PK, et al., Genome Med, 2014)
- Identifying biomarkers for improved classification of COPD patients (Gomez-Cabrero, D. et al., J. Transl Med, 2014).
Biomarker discovery for ecotoxicology
- Gene transcription, metabolite and lipid profiling were used to identify biomarkers in Daphnia magna predictive of exposure to flame-retardants. (Scanlan, LD., et al. Environ Sci Technol, 2015)
- Hepatic stickleback transcriptomics and metabolomics data were used to determine biomarkers for dibenzanthracene contamination (Williams TD, et al., Environ Sci Technol, 2009) and aquatic levels of ethinyl-estradiol. (Katsiadaki I et al., Aquat Toxicol. 2010).
Back to: Computational Biology Facility