The biochemical stratification of amyotrophic lateral sclerosis

Motor neuron disease (MND, also known as amyotrophic lateral sclerosis, ALS) is a fatal, incurable disease that causes progressive muscle weakness. MND affects people in very different ways: weakness can begin first in any body part, sometimes people have symptoms affecting their thoughts and behaviours as well as weakness, and, although people with MND live on average for less than three years after weakness develops, some will live for over a decade. Why MND behaves in these different ways is not understood, but it is likely that it relates to differences in the way that the disease affects cells in the nervous system. These differences are probably a major reason why a large number of drug trials for MND have failed.

My research will study the molecules (called proteins) that perform virtually all of the tasks in every cell of the body, including nerve cells. I will examine the different patterns of proteins found in the spinal fluid, reflecting changes in nervous system cells, of a large number of MND patients to try to understand what is happening in the nervous system to account for the differences in the ways that MND affects people. Through an advanced computer technique called machine learning, I aim to identify protein patterns that relate to different patterns of MND. Using these protein patterns, I hope that in five years' time this will enable drug trials to be more specifically tailored to a person's MND, improving the likelihood of drug trial success.

I will perform deep proteomic profiling using unbiased data independent acquisition mass spectrometry proteomics (Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS), ~2000 proteins depth). The samples will come from Oxford MND Centre's existing and growing ALS CSF collection (n=169 participants, with some longitudinal visits, total n~250). There are over 50 healthy control CSF samples for comparison and a smaller set of non-ALS motor weakness samples. All samples come from well-clinically-characterised patients, and are accompanied by key established biomarker data such as neurofilament and chitinase.

Classical statistics such as t-tests and regression will be applied to the data to detect relationships between clinical parameters and CSF protein dysregulation in ALS patients versus controls. This univariate analysis may itself lead to novel findings for further validation and pathway analysis, as it will be the largest and highest depth integrated analysis performed on ALS CSF to date. I will then apply more complex statistical techniques such as machine learning to detect the presence of patient clusters by grouping similar data points together, highlighting underlying patterns, and reducing the high dimensionality of the data. This will involve randomly partitioning the data into discovery (2/3) and holdout datasets (1/3), using the discovery set for model training and tuning, and then applying the models to the holdout set to prospectively test internal validity.

I will then explore the biologic relevance of cluster membership through pathway analysis, weighted gene correlation network analysis, gene ontology and cell type enrichment analysis. The Oxford MND group has the multidisciplinary expertise to orthogonally study the most relevant findings. Finally, the key indicator proteins associated with ALS subtype will be incorporated into a target proteomics array for a cost-effective prospective trial at the Oxford MND clinic.