An interpretable machine learning-based cerebrospinal fluid proteomics clock for predicting age reveals novel...

Area of Science:

• Neuroscience
• Biochemistry
• Computational Biology

Background:

• Biological age can be predicted using machine learning "biologic clocks."
• Organs and biofluids may age differently than the whole organism.
• Existing clocks using plasma or tissue samples have limitations for brain aging assessment.

Purpose of the Study:

• To develop a machine learning clock using cerebrospinal fluid (CSF) proteomics to predict brain aging.
• To identify novel proteins and pathways involved in central nervous system (CNS) aging.
• To inform the development of brain aging-related disease mechanisms and anti-aging therapeutic targets.

Main Methods:

• Developed a machine learning model utilizing CSF proteomic data to predict chronological age.
• Validated the clock's accuracy using Pearson correlation and Mean Estimated Error (MAE).
• Analyzed highly weighted proteins and developed a minimal 109-feature protein clock.

Main Results:

• The CSF proteomic clock achieved a 0.79 Pearson correlation and 4.30-year MAE in the validation cohort.
• A minimal clock with 109 protein features demonstrated similar accuracy (0.75 correlation, 5.41-year MAE).
• Identified novel proteins predictive of age through interactions, not direct correlation.

Conclusions:

• The CSF protein aging clock accurately predicts biological age and offers insights into CNS aging.
• This approach identifies novel age-related proteins and pathways not apparent through individual analysis.
• The clock serves as a valuable tool for understanding brain aging and developing targeted therapies.