Reevaluating principal component analysis in geroscience: A call for nonlinear approaches in AI-based evaluations

Area of Science:

• Gerontology
• Artificial Intelligence
• Biostatistics

Background:

• Explainable AI is crucial for evaluating aging and longevity interventions.
• Principal Component Analysis (PCA) is a commonly used method.
• PCA's linear and parametric nature may not suit complex, nonlinear geroscience data.

Purpose of the Study:

• To address the limitations of Principal Component Analysis (PCA) in analyzing gerontology data.
• To propose alternative statistical methods for more accurate intervention evaluation.
• To highlight the importance of nonlinear and nonparametric approaches in aging research.

Main Methods:

• Critique of Principal Component Analysis (PCA) for nonlinear biological data.
• Advocacy for nonlinear and nonparametric statistical methods.
• Examples of recommended methods: Spearman's rank correlation and Kendall's tau.

Main Results:

• PCA can misrepresent complex, nonlinear relationships common in aging research.
• Reliance on PCA may obscure crucial biological insights.
• Misinterpretations of intervention effects are possible with linear methods.

Conclusions:

• Nonlinear and nonparametric methods offer enhanced accuracy for analyzing aging intervention data.
• Adopting methods like Spearman's rank correlation and Kendall's tau can improve geroscience research.
• Revising methodological approaches is key to informed evaluations in aging and longevity studies.