Simple questions about neurodegenerative disease: Where? What?

Alzheimer's disease is largely caused by genetics rather than lifestyle factors. The probability of an identical twin developing the disease if their co-twin already has it is ~79%. By comparing the DNA of people who develop a disease against healthy people, we can identify sites in their DNA that increase the likelihood of getting the disease. In recent years it has been established that thousands of changes to the human genome can contribute to disease risk. Each variant explains only a tiny part of the risk, but collectively these variants account for much of an individual's susceptibility. Biological processes implicated by these loci can be understood as being causally involved in the disease. Understanding the mechanisms of neurodegenerative diseases has thus become a statistical problem: we need to find the patterns that link the variants together. There are many open questions about neurodegenerative diseases. For most brain diseases, we still lack answers to seemingly basic questions: which part of the brain has gone wrong, and what happened there? A core focus of my work has been using genetics to answer the first of these: which cell types cause particular diseases. The basic premise is that genes which cause a disease will be disproportionately utilised (expressed) by the causal cell type. Using this approach, I showed that schizophrenia specifically affects certain neuron subtypes—ruling out some major theories of the disease’s aetiology. Similarly, I published the first paper showing that Alzheimer’s genes are enriched in microglia, the brain’s immune cells. This surprised the field, which had long considered Alzheimer’s a neuronal disease. Contrary to expectation, no enrichment of Alzheimer's risk genes has been found in neurons. I then applied the method to Parkinson’s: confirming enrichment in dopaminergic neurons, but also implicating oligodendrocytes—previously unlinked to the disease. To resolve finer questions (e.g., whether Alzheimer’s risk is specific to brain microglia or shared with all macrophages), we need higher-resolution methods: the first part of this project aims to deliver these. What we then really need to know is: what happens within a cell to cause the disease? Once we understand this, we can develop drugs to reverse the process. Again, we can use statistics—if we gather enough data on molecular processes in the disease-causing cell types. Most disease-causing genetic variants act by disrupting interactions between DNA and proteins. DNA-binding sites tend to have distinct sequences, so we can train machine learning models to predict them and simulate how genetic variation alters them. We can then test whether specific molecular interactions (e.g., binding of a transcription factor) are disrupted. Once identified, we can begin efforts to reverse these disease-linked disruptions. This project will generate the required data and develop cutting-edge machine learning models to put this approach to the test.