Plant phenotyping is crucial for breeding improved plant varieties to meet growing population needs. However, the data generated is complex due to diverse experimental parameters and settings, hindering reuse. The sheer volume and heterogeneity of plant phenotyping datasets, coupled with poor documentation and varying data types, pose significant challenges to meta-analyses. This complexity manifests in both epistemological (ambiguities, missing documentation) and logistical (data undiscoverability, incompatible data types) obstacles. To address these issues, the FAIR data principles provide a framework to improve data management and facilitate data integration and reuse across distributed resources. The MIAPPE metadata standard aids findability, interoperability, and reusability, while accessibility is addressed through implementations such as the Breeding API. Data reuse without FAIR principles and supporting standards becomes a laborious process often described as "data archaeology," requiring extensive effort to reconcile and integrate fragmented information from multiple sources. This paper investigates how FAIR principles and MIAPPE can alleviate these challenges and improve the efficiency of data reuse in plant phenotyping.

The authors reference several key works highlighting the challenges of plant phenotyping data reuse and the potential of FAIR principles and MIAPPE for improvement. Coppens et al. (2017) and Pieruschka & Schurr (2019) emphasize the need for improved data integration and data-driven approaches to unlock the potential of plant phenotyping data. Wilkinson et al. (2016) introduced the FAIR Guiding Principles, which this study uses as a framework. Papoutsoglou et al. (2020) explored the implementation of MIAPPE 1.1 for enhancing the reusability of plant phenomic datasets. Selby et al. (2019) discuss the BrAPI, an application programming interface for plant breeding applications which contributes to accessibility. Hurtado-Lopez's (2012) doctoral thesis, which serves as the case study for this research, demonstrated successful data reuse despite logistical challenges, highlighting the need for better data practices and documentation – three key elements being content, origin/source, and structure, all of which are addressed by MIAPPE.

This research takes the form of a proof-of-concept (PoC) implementation that demonstrates how FAIR principles and community standards can be applied to facilitate data reuse in a meta-analysis of potato developmental traits. The PoC re-implements a meta-analysis from Hurtado-Lopez's doctoral thesis, investigating genotype by environment and QTL by environment interactions across five experiments conducted in four different locations over eleven years. The methodology involves six key steps: (a) locating relevant phenotypic data using a FAIR Data Point (FDP); (b) exploring investigation and study metadata; (c) verifying genotype overlap across experiments; (d) finding temporally and spatially aligned weather data; (e) reusing traits per genotype; and (f) aggregating traits with weather data. The datasets used are from Hurtado-Lopez's thesis, including five phenotypic experiments and associated photoperiod and temperature data. The authors assume the role of a researcher undertaking a multi-environment study. The MIAPPE standard guides metadata creation, and the data is represented in RDF format and made available through a SPARQL endpoint. A custom FDP is constructed for this case, addressing a limitation of the FDP specification by embedding MIAPPE metadata into the dataset level for improved searchability and findability. The processes are documented using Jupyter notebooks. Weather data was acquired from both the original experimental data files and through online sources. The authors describe in detail the challenges and benefits of the approach. Specifically, they address how the FAIR principles are applied at each step of data discovery, acquisition, and integration, with details on metadata standardization, data formatting, and data integration. The implementation relies on MIAPPE metadata, RDF data model, SPARQL for querying the data, and the FDP infrastructure, showcasing a transparent pipeline that combines heterogeneous data across institutional and domain silos.

The PoC implementation successfully demonstrates the feasibility of reusing plant phenotyping data using FAIR principles and MIAPPE. The authors show how relevant phenotypic datasets can be discovered and explored using the FDP infrastructure. They illustrate how metadata can be used to verify genotype overlap across experiments and to locate matching weather data. The integration of phenotypic and environmental data allows for exploratory analyses, generating visualizations demonstrating the correlation between temperature, photoperiod, and tuber weight for different genotypes. The analysis revealed sharp differences in genotype performance across experiments and substantial variability even within the same genotype across different environments. A visualization combined phenotypic, experimental, and environmental data, plotting a range of tuber weight performance (minimum and maximum) for each genotype across different cumulative photo-beta thermal time (PBTT) values. This visualization aids in understanding genotype performance stability across environments. While the analysis didn't lead to new biological insights, it effectively demonstrated the value of the FAIR approach for exploratory data analysis. The process highlighted the critical role of comprehensive and unambiguous metadata for efficient data reuse. Challenges included the interpretation of existing metadata, which was often in free-text format, necessitating communication with the original data collectors. Data cleaning and harmonization also proved time-consuming. The process took on average two weeks per experiment, emphasizing that structured documentation such as MIAPPE greatly improves efficiency.

This work underscores the significant benefits of applying FAIR principles and standards like MIAPPE to enhance data reusability in plant phenotyping. The PoC successfully demonstrates a pipeline incorporating all aspects of FAIR, enabling efficient data discovery, acquisition, analysis, and integration. The challenges encountered during the PoC implementation highlight the importance of creating metadata that is easily understood by both machines and humans. The current FDP specification's limitations regarding dataset content description were addressed by integrating MIAPPE metadata directly into the dataset level, improving search functionality and enabling content-oriented searches. While the PoC demonstrates the effectiveness of FAIRification, the high technical knowledge barrier necessitates improved user experience with graphical user interfaces. The study strongly advocates for adopting FAIR practices, even with existing technical limitations, prioritizing data with good documentation and identifiers, encouraging a balance between FAIR-ready datasets and fully developed technical infrastructures.

This study successfully demonstrates the benefits and challenges of applying FAIR principles to existing plant phenotyping data. Using a case study and a proof-of-concept implementation, the authors highlight how FAIRification improves data reuse and the process's efficiency. The study advocates for community-wide adoption of FAIR principles and standards, stressing the need for comprehensive metadata and user-friendly interfaces. Further research should focus on automating FAIRification processes and improving user experience.

The study's main limitation is the single proof-of-concept implementation, precluding generalized conclusions on certain FAIR-related issues, such as retrospective versus prospective FAIRification and automation potential. While the authors provide informed opinions, further research is needed to conclusively assess these aspects. Another limitation concerns the scope of the meta-analysis which is limited to a selection of traits and experiments in the initial study. A more comprehensive evaluation might further refine the insights.