LongDat: an R package for covariate-sensitive longitudinal analysis of high-dimensional data

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Longitudinal data analysis is crucial for understanding biological processes over time.
• Identifying direct vs. indirect effects and mechanistic intermediates in complex datasets remains challenging.
• Existing tools often struggle with high-dimensional covariate data in longitudinal studies.

Purpose of the Study:

• To introduce LongDat, a novel R package for analyzing longitudinal multivariable data.
• To enable simultaneous accounting for a large number of covariates.
• To facilitate the differentiation of direct and indirect effects of interventions and identify mechanistic intermediates.

Main Methods:

• LongDat employs advanced statistical methods tailored for longitudinal data.
• The package is designed to handle multiple types of data, including microbiome, binary, categorical, and continuous.
• Performance was evaluated against established tools (MaAsLin2, ANCOM, lgpr, ZIBR) using simulated and real-world data.

Main Results:

• LongDat demonstrated superior accuracy, runtime efficiency, and lower memory usage compared to existing methods.
• The package excels particularly in scenarios with a high number of covariates.
• Robust biomarker discovery in high-dimensional longitudinal datasets is facilitated by LongDat.

Conclusions:

• LongDat is a computationally efficient and memory-sparing R package for longitudinal data analysis.
• It provides a powerful tool for dissecting complex intervention effects and identifying key biological intermediates.
• The package is suitable for a wide range of applications, including microbiome research and other high-dimensional biological data.