Learning Gaussian Graphical Models from Correlated Data.

Area of Science:

• Biostatistics
• Genomics
• Network Analysis

Background:

• Gaussian Graphical Models (GGMs) are crucial for analyzing complex relationships in biomedical research.
• Standard GGM inference methods assume independent observations, which is often violated in clustered and longitudinal data.
• Ignoring correlation in data can lead to inflated Type I errors, compromising study findings.

Purpose of the Study:

• To propose a novel Bootstrap algorithm for inferring GGMs from correlated data.
• To address the limitations of existing methods when dealing with non-independent observations in biomedical studies.
• To provide a statistically robust approach for analyzing complex relationships in family-based and longitudinal datasets.

Main Methods:

• A Bootstrap algorithm was developed to estimate GGMs from correlated observations.
• Extensive simulations using correlated data from family-based studies were conducted to evaluate the method.
• The proposed method was applied to analyze Polygenic Risk Scores from the Long Life Family Study.

Main Results:

• The Bootstrap method effectively controls Type I error inflation in correlated data.
• The proposed algorithm retains statistical power compared to alternative methods that ignore correlation.
• Application to the Long Life Family Study demonstrated robust Type I error control in a real-world dataset.

Conclusions:

• The proposed Bootstrap algorithm provides a reliable method for inferring GGMs from correlated biomedical data.
• This approach is suitable for complex family-based and longitudinal studies, improving the accuracy of relationship inference.
• The method offers a significant advancement for analyzing high-dimensional correlated data in genetic and biomedical research.