Putting the dynamic pathosome in practice: a novel way of analyzing longitudinal data

TL;DR

This paper introduces a functional linear model approach inspired by the dynamic pathosome concept to analyze longitudinal phenotypic data, demonstrating improved variance explanation and identifying key developmental time points.

Contribution

It presents a novel functional linear modeling method based on the dynamic pathosome concept for analyzing longitudinal data, outperforming classical models.

Findings

Functional linear models explain more variance in age at menarche than classical models.

Identified crucial time points that approximate functional model performance.

Growth trajectories predict 97% of height variance at 18 years.

Abstract

Previously we have developed the concept of the dynamic pathosome, which suggests that individual patterns of phenotype development, i.e., phenotypic trajectories, contain more information than is commonly appreciated and that a phenotype's past trajectory predicts its future development. In this article, we present a pathosome-inspired approach to analyzing longitudinal data by functional linear models. We demonstrate how to use this approach and compare it with classical linear models on data from the Czech section of the European Longitudinal Study of Pregnancy and Childhood (ELSPAC). Our results show that functional linear models explain more observed variance in age at menarche from height and weight data than the commonly used approaches. Furthermore, we demonstrate that functional linear models can be used to identify crucial time points that can be used to create linear models achieving almost the same performance as functional linear models. In addition, we use data from the Berkeley growth study (BGS) to demonstrate that growth trajectories from birth to 15 years can be used to explain 97% of observed variance of height at 18 years, thus supporting the notion that a phenotype's past trajectory affects its future course. Overall, this article presents experimental support for the concept of the dynamic pathosome and presents a method that can be used as a powerful tool for analyzing quantitative longitudinal data.