The role of human mitochondria in complex disease risk
Over recent years, mitochondria have more frequently been implicated in common complex diseases, particularly those linked to ageing. For example, increased mitochondrial DNA damage and oxidative phosphorylation defects are observed in coronary artery disease patients, hyperglycemia is thought to trigger diabetic complications through mitochondria derived reactive oxygen species, and shifting energetic profiles, utilising glycolysis over oxidative pathways, is key in many types of cancer. Despite these links to a range of cardiovascular, metabolic and neurological disorders, it is often not known whether altered mitochondrial processes play a causal role in the risk of developing disease, or occur as a consequence of them, making it hard to develop targeted treatment strategies.
Mitochondria contain their own genome, coding for 15 key protein coding genes that all form part of the electron transport chain (ETC), required for cellular energy production via oxidative phosphorylation. However, all other proteins that make up the ETC, as well as all other proteins that are involved in the replication, repair, transcription and translation of mitochondrial products are encoded in nuclear DNA. As such, interactions between these two genomes are vitally important for correct cell function, and perturbations in these relationships have been linked to several common complex diseases.
The aim of this project is to apply machine learning techniques to multi-dimensional nuclear and mitochondrial data in order to predict mitochondrial processes across a large number of individuals, and test the impact of altered mitochondrial phenotypes on disease risk. Within this, we will build models of mitochondrial transcriptome features such as gene expression and RNA modification using large population based RNA sequencing datasets from both healthy and diseased individuals (focussing on cardiovascular and neurological disorders) and then integrate these data with high coverage genetic information and machine learning techniques (using shrinkage methods such as LASSO, ridge and elastic net regression, but also more complex techniques such as neural networks) to predict these events in large cohorts (such as UK Biobank). By comparing predicted mitochondrial events to disease related traits (including those implicated in cardiovascular and neurological diseases, such as lipid traits), we aim to identify the genetic and molecular mechanisms linking mitochondria to common disease.
Numerous population level RNA sequencing datasets (GTEx, CARTaGENE, TwinsUK), single cell RNA sequencing datasets and large scale genetic data (e.g. UK Biobank). Approvals granted for use in this work.
Mitochondria, Complex Disease, Transcription, Machine Learning, Genomics