ALS molecular subtypes are a combination of cellular and pathological features learned by deep multiomics classifiers

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Amyotrophic lateral sclerosis (ALS) presents with significant genetic and clinical heterogeneity.
• Previous transcriptomic studies identified potential ALS subtypes linked to mitochondrial dysfunction (ALS-Ox), neuroinflammation (ALS-Glia), and TDP-43 pathology (ALS-TE).

Purpose of the Study:

• To develop and apply a deep neural network classifier, DANCer, for precise identification of ALS molecular subtypes.
• To investigate the correlation of these subtypes with clinical features, specifically disease duration.
• To analyze cellular and molecular alterations within identified subtypes.

Main Methods:

• Development of a deep neural network classifier (DANCer) for ALS molecular subtyping.
• Application of DANCer to an expanded cohort from the NYGC ALS Consortium.
• Analysis of single-nucleus transcriptomes to examine cell-type-specific alterations.

Main Results:

• DANCer successfully classified ALS samples into molecular subtypes.
• Two subtypes, ALS-TE (in cortex) and ALS-Glia (in spinal cord), showed a strong correlation with disease duration.
• Single-nucleus RNA sequencing revealed both common and subtype-specific transcriptomic changes across neurons and glia.

Conclusions:

• ALS molecular subtypes are defined by distinct transcriptomic signatures and cellular pathology.
• These subtypes, particularly ALS-TE and ALS-Glia, are associated with key clinical parameters like disease duration.
• Understanding these subtypes provides insights into the complex cellular mechanisms underlying ALS progression.