Abstract 1
Category: Basic Science
At the end of the session, participants will be able to:
- How AI can be used as a research tool to understand the comprehensive biological underpinnings of neurodegenerative disease
COI Disclosure:
None to disclose
Presenter
Dr. Ain Kim is a post-doctoral researcher at the Tanz Centre for Research in Neurodegenerative Diseases (CRND), University of Toronto. Dr. Kim’s research explores the application of artificial intelligence (AI) to transform and deepen our understanding of neurodegenerative diseases, with a particular emphasis on the morphological, biochemical, and cytopathological characterization of these diseases.
Ain Kim1, Shelley L. Forrest1,2, Gabor G. Kovacs1,2,3,4,5,6,7
1Tanz Center for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Canada
2Krembil Brain Institute, University Health Network, Toronto, ON, Canada
3Rossy Centre for PSP, Toronto Western Hospital, Toronto, Canada
4Edmond J. Safra Program in Parkinson’s Disease and the Morton and Gloria Shulman Movement Disorders Clinic, Toronto Western Hospital, Toronto, Canada
5Division of Neurology, University of Toronto, Toronto, Canada
6Laboratory Medicine Program, University Health Network, Toronto, Canada
7Department of Laboratory Medicine & Pathobiology, University of Toronto, Toronto, Canada
Target Audience:
Pathologists, Other – Write In (Required): researchers
CanMEDS:
Professional
Decoding the Diseased Brain: Using Machine Learning to Gain Comprehensive Understanding of Neurodgenerative Diseases
Abstract
Introduction: Neurodgenerative diseases are characterized by neuronal loss and deposition of misfolded proteins in the human brain. Since neuropathological evaluation remains the gold standard for definitive diagnosis, many efforts have been made to understand the complex neuropathological and molecular changes underlying neurodegenerative diseases using fully characterized post-mortem tissues. These include evaluating cytopathologies, morphologies of pathological inclusions, biochemical properties of misfolded proteins using seed amplification assays (SAA), and genotypic information. However, the results of these evaluations are not independent, but rather should be interpreted together. In order to integrate results from multiple disciplines and achieve a comprehensive understanding of the disease, a powerful and scalable tool is needed.
Methods: We used machine learning, a branch of artificial intelligence, to learn latent patterns from high dimensional data and incorporate results from multiple disciplines.
Results: Machine learning-based analysis has allowed: i) the discovery of novel subtypes of multiple system atrophy based on morphological variables, ii) the integration of alpha-synuclein cytopathology and seeding behavior to describe cell type-specific seeding in synucleinopathies, iii) the understanding of differences in in inflammatory markers of distinct inheritable DNA sequence patterns, iv) the validation of subtype and staging in progressive supranuclear palsy, and v) the prediction of 4R-tauopathies using morphological differences in coiled bodies.
Conclusion: Machine learning can be applied not only for diagnostic predictions in clinical settings, but also as a research tool to understand the comprehensive biological underpinnings of neurodegenerative disease.