A machine learning approach to spike sorting to reveal neuronal signalling defects in models of neurodegeneration

The brain comprises a complex network of interconnected neurons. This network is significantly compromised in neurodegenerative diseases such as Alzheimer’s, Parkinson’s, and Niemann-Pick type C disease. The current understanding of the molecular mechanisms underlying these disorders is limited. Pathological protein aggregation is believed to impact neuronal signalling by disrupting ion channel activity and neurotransmitter release. These alterations compromise synaptic transmission and neuronal communication, which are crucial for maintaining cognitive and motor functions. However, the limited resolution and efficacy of current neurophysiological analysis techniques hinder a complete molecular understanding of neurological disorders. This particularly includes their initiation and propagation, ultimately resulting in the current lack of treatments. Extracellular recordings of neuronal activity constitute a powerful tool for investigating the dynamics of neural networks and the activity of individual neurons. Microelectrode arrays (MEAs) allow for recordings with a high electrode count, generating extensive datasets of neuronal information. Furthermore, MEAs capture extracellular field potentials from cultured cells, resulting in highly complex neuronal signals, requiring precise spike sorting for meaningful data extraction. Nevertheless, conventional spike sorting methods face limitations in recognising diverse spike shapes, thereby constraining the full utilisation of the rich dataset acquired from MEA recordings. To overcome these limitations, a machine learning-based spike sorting approach is developed. Multiple algorithms have been developed and evaluated, and a final algorithm, PseudoSorter, is presented. PseudoSorter is a spike sorter based on advanced self-supervised learning techniques, a distinctive density-based pseudo-labelling strategy, and an iterative fine-tuning process to enhance spike sorting accuracy. Through extensive benchmarking on large-scale simulated datasets, the superior performance of the developed spike sorting approaches compared to existing alternatives is demonstrated. Several applications of machine learning-based spike sorting to investigate neuronal signalling defects in models of neurodegenerative diseases have been explored. Firstly, graphene MEA recordings and super-resolution imaging techniques reveal critical insights into the structural and functional alterations of neurons in a model of Niemann-Pick type C disease. The presented findings demonstrate a loss of neuronal synchronicity and structural impairments, providing a deeper understanding of the disease’s impact at the cellular level. Secondly, MEA recordings from hippocampal neurons were exposed to subneuronal concentrations of monomeric Tau, a protein associated with Alzheimer’s disease (AD). The presented results, validated against patch clamp experiments, unveil that monomeric Tau at subneuronal concentrations induces stimulation-dependent disruptions in both local and global activity of hippocampal neurons. Finally, electrophysiological MEA recordings were combined with calcium imaging to investigate the effects of cholesterol and its metabolites in the presence or absence of α-synuclein, a protein related to Parkinson’s disease. The results indicate that treatments involving 27-hydroxycholesterol in the presence or absence of α-synuclein lead to significant neuronal defects, including reduced electrical and calcium activity and synchronisation. The developed machine learning-based spike sorting methods represent versatile analysis tools for neuroscientists across diverse research domains. Potential applications span from fundamental research of neural network dynamics to practical investigations into neurodegenerative disease models and the exploration of brain-computer interfaces. The provided analysis tools can generate more detailed insights into the molecular mechanisms underlying neurodegenerative diseases, potentially opening up new avenues for testing novel therapeutic strategies for neurodegenerative diseases.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2024

Author dc:creator

• Brockhoff, Marius

Advisor dc:contributor.advisor

• Kaminski Schierle, Gabriele

• alzheimer's disease
• machine learning
• neurodegeneration
• spike sorting

Licence

• https://www.repository.cam.ac.uk/bitstreams/8d993940-42b9-41d7-808c-d915ae7dcb28/download
• http://purl.org/NET/rdflicense/allrightsreserved

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.120744

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/388346