Shared Functional Representation and Genomic Region Association Analysis of Longitudinal Phenotypes in Rice.

Abstract

High-throughput phenotyping technologies enable plant growth to be measured repeatedly at high temporal resolution, providing rich information for studying the dynamic genetic basis of longitudinal traits. However, growth trajectories from repeated experiments contain both reproducible variation among lines and experiment-specific variation. In this study, we analyzed projected shoot area data from 356 rice lines measured in three independent experiments using an integrated genomic region association framework based on shared functional representation. We extracted the major longitudinal variation shared across experiments and characterized the covariance eigenspectra using spectral entropy and entropy-based effective rank. The first shared functional principal component explained 96.069% of shared variation, while the shared component had a normalized spectral entropy of 0.101 and an effective rank of 1.198. Regional association analysis using 200 kb genomic windows, with adjustment for population structure and leave-one-chromosome-out genetic relatedness, identified a region at 14.2-14.4 Mb on chromosome 10 that remained significant after false discovery rate correction (q=0.004). Further analyses showed time-varying regional effects and differences in association evidence across experiments, while the chromosome 10 signal was attenuated in an endpoint benchmark. Our analysis provides a method for linking longitudinal phenotypic variation shared across repeated experiments with localized genomic variation.