Hyperbolic topological data analysis mapper reveals dynamic trait–environment patterns in plant phenomics

Abstract

Modern plant phenotyping faces the challenge of interpreting complex, high-dimensional data. Traditional analytical tools often fail to capture the non-linear, hierarchical, and temporal relationships that define plant responses under multifactorial conditions. We present the Hyperbolic Topological Data Analysis Mapper (HTDA-Mapper), a novel algorithm designed to overcome these limitations by embedding data in Poincaré ball space. Unlike conventional Euclidean approaches, HTDA-Mapper preserves the hierarchical structure of phenotypic traits, improves cluster resolution, and reveals hidden growth trajectories across treatments and time, offering a powerful means to explore latent phenoms. The pipeline supports both quantitative data and images. When integrated with unsupervised contrastive learning, HTDA-Mapper identifies similarities and differences in raw image data without requiring manual labelling or post hoc processing. We applied this framework to a high-throughput phenotyping (HTP) dataset of over 27,000 images of Arabidopsis thaliana seedlings exposed to varying nutrient levels and priming agents at different concentrations over seven days. Using cubical complexes, HTDA-Mapper mapped relationships between treatment variables, compound concentrations, and phenotypic outcomes. Furthermore, it reliably detected compound-specific effects, uncovered dynamic trait–environment interactions, revealed phenotypic trajectories not captured by conventional methods, and facilitated biologically meaningful interpretation of the complex dataset. By preserving the geometry and temporal evolution of plant development, HTDA-Mapper sets a new standard for HTP analysis. Beyond phenomics, it is a versatile tool for other omics, such as transcriptomics and metabolomics, where structured, high-dimensional data is prevalent. HTDA-Mapper can accelerate data-driven crop improvement by uncovering effective compounds, robust genotypes, and adaptive growth strategies that enhance plant resilience.

Plant phenotyping relevance

植物フェノミクスの高次元画像・形質データを解析するHTDA-Mapperアルゴリズムを開発し、27,000枚超の植物画像データで適用・評価しているため、解析手法が中心的である。

Code and data availability

保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。