Musings from The Power List: Gerrit Renner

Gerrit Renner explains why an analytical measurement does not end at the detector – and why trustworthy AI will depend on transparent, traceable data

Category	Detail
Condition	Analytical data science and trustworthy analytical measurements
Key Mechanisms	Data-processing and chemometric approaches that retain uncertainty, quality, and provenance across analytical workflows.
Target Population	Researchers and practitioners in analytical science, including those working with complex mass-spectrometric data.
Care Setting	Analytical research workflows, from signal recording through processing and scientific interpretation.

Data processing is described as an integral part of analytical measurement, not a separate post-measurement step.
Processing steps such as centroiding, feature detection, integration, and modeling introduce assumptions and uncertainty.
The research aims to make processing transparent and quantitative, including propagation of uncertainty and quality measures.
Machine-actionable analytical data should carry metadata, provenance, uncertainty, and semantic meaning.
FAIR data practices are advocated throughout the analytical workflow, beginning at acquisition.

Renner describes developing approaches that make processing steps transparent and quantitative, including uncertainty estimates and quality measures that can be propagated through analytical workflows.
Renner argues that analytical workflows should preserve FAIR, machine-actionable data and its context from acquisition onward.

Record processing decisions and uncertainty across transformations from detector signal to final result, as advocated by Renner.
Retain metadata and provenance to support interpretation and reconstruction of analytical results.

Renner notes that preprocessing, peak detection, integration, and filtering decisions can substantially affect outcomes.
Renner warns that context, provenance, and processing information may be lost when FAIR practices are applied only after data generation and processing.
The source says context may be hidden in proprietary formats, software settings, or scientists’ notebooks, and processing decisions can be difficult to reconstruct.

Not applicable; the article concerns analytical science and data workflows, not patient care.

No treatments or prescribing information are discussed.

Renner describes data structures, algorithms, statistics, and scientific interpretation as parts of the same measurement chain.
Renner argues that FAIR and machine-actionable data should be maintained throughout the workflow, beginning at acquisition.
Preserve information about how analytical values were produced and their uncertainty and quality, consistent with Renner’s stated research focus.