Close
Supervisors:
Prof Shamima Rahman and Dr Konrad Wagstyl
Project Description:
Background:
Primary mitochondrial diseases (PMDs) are clinically, biochemically and genetically heterogeneous conditions unified by disturbances in mitochondrial function [1]. Clinical features are typically multisystemic in childhood and frequently involve the brain. Neurological features include motor and cognitive developmental delays, dystonia, ataxia and seizures [2]. Exome and genome sequencing have transformed mitochondrial medicine, with approaching 400 different genes across two genomes now linked to PMDs [3]. However many important questions remain unanswered, including the reasons for tissue specificity and variability of clinical presentation of individuals sharing identical gene defects, and a lack of disease-modifying therapies and biomarkers to monitor disease progression and/or response to treatment.
Aims/Objectives:
Investigate genetic and neuroimaging markers in PMD using multi-omics analysis:
1. Use neuroimaging and machine learning to identify biomarkers that can be used to aid diagnosis, quantify the severity of disease burden and ultimately track treatment response in clinical trials.
2. Use multimodal genetic data to identify regional, cellular, subcellular and functional characteristics of genes associated with PMD. We will analyse these features to better understand the common biological pathways associated with PMD and stratify potential novel risk genes.
3. Identify potential correlations between genetic mutations and neuroimaging markers, to improve diagnosis and understanding of disease mechanisms.
Methods:
The project will involve retrospective analysis of MRI data from a large cohort of patients with PMD scanned at Great Ormond Street Hospital. Structural T1 and T2 MRI neuroanatomical data will be compared with normative developmental growth charts to identify individualised markers of disrupted neurodevelopment. Machine learning techniques, developed at GOSH for identifying subtle epileptogenic abnormalities [4,5], will be adapted to localise and quantify lesion burden in children with PMDs affecting the brain. Openly available atlases of regional, cellular and sub-cellular gene expression data will be used to identify shared characteristics of rare genetic mutations associated with PMD. Regional brain patterning will be compared with neuroanatomical disruption to characterise the biological processes underpinning imaging changes. An AI model will be trained to classify PMD risk genes according to these features and used to predict the likelihood of disease association across the genome.
Timeline:
Year 1 – Development of neuroimaging and machine learning biomarkers in genetically-confirmed cases with PMD;
Year 2 – Integration of neuroimaging and genomic datasets; identification of genotype-imaging correlations and hypothesis generation; data analysis of validation cohort to confirm genotype-phenotype correlations
Year 3 – Completion of data analysis; thesis writing, papers
References:
1. Rahman and Rahman (2018) Mitochondrial medicine in the omics era. Lancet 391(10139):2560-2574. PMID: 29903433.
2. Rahman (2020) Mitochondrial diseases in children. J Intern Med 287(6):609-633. PMID: 32176382
3. Schlieben and Prokisch (2023) Genetics of mitochondrial diseases: Current approaches for the molecular diagnosis. Handb Clin Neurol 194:141-165. PMID: 36813310.
4. Spitzer et al (2022) Interpretable surface-based detection of focal cortical dysplasias: a Multi-centre Epilepsy Lesion Detection study. Brain 145(11):3859-3871. PMID: 35953082.
5. Eriksson et al (2023) Predicting seizure outcome after epilepsy surgery: Do we need more complex models, larger samples, or better data? Epilepsia PMID: 37129087.
Contact Information:
Prof Shamima Rahman