Musings from The Power List: Ansgar Korf

Could reusable mass spectrometry data transform analytical science?

October 9, 2026
7 min

The Analytical Scientist

Overview

Ansgar Korf describes developing computational methods and open infrastructure through MZmine to help researchers process and reuse untargeted mass-spectrometry data. He identifies data comparability, workflow quality, and recognition of analytical science as key challenges, and describes FAIR-MS as an effort to make processed data findable and reusable for machine learning.

Background

Untargeted mass spectrometry generates extensive small-molecule data, while interpretation and software handling can consume substantial researcher time, according to Korf. He argues that data from different laboratories and instruments are often insufficiently comparable for large-scale machine-learning applications. He also emphasizes that study design, sampling, preparation, acquisition, and processing jointly constrain analytical results. Related clinical-metabolomics guidance identifies preanalytical variation, inter-laboratory reproducibility, batch effects, and qualification as barriers to clinical reliability.

Data Highlights

The interview provides no numerical trial results. It reports that Korf conceived the idea behind mzio in 2020 and that MZmine 3 was published in Nature Biotechnology before the interview describes events converging by 2023.

Key Findings

Korf’s team develops computational methods and open infrastructure through MZmine to reduce data-processing bottlenecks in untargeted small-molecule mass spectrometry.
He describes FAIR-MS as applying MZmine processing capabilities to make mass-spectrometry data findable and reusable at a scale relevant to machine learning.
Korf expects untargeted measurement to become more common as software capabilities improve, while noting that targeted methods remain important when a validated answer is required.
He cautions that systematic workflow errors—including matrix effects, ion suppression, batch drift, and sample degradation—may be learned by models rather than removed by increasing data scale.
Korf characterizes the status of analytical science as a recognition problem: he says method development and analytical expertise are undervalued and software is often funded as a research byproduct.
A 2026 review of clinical metabolomics identifies nonuniform preanalytics, inter-laboratory reproducibility, batch effects, and lack of regulatory qualification as barriers; a 2026 international survey of 85 laboratories in 27 countries reports QA/QC practices and recommendations for GC-MS and GC×GC-MS metabolomics.

Clinical Implications

The interview does not establish clinical effectiveness for MZmine, FAIR-MS, or AI-based interpretation. The cited clinical-metabolomics review and CCQM report identify harmonized workflows, QA/QC, reference materials or methods, and validation approaches as needs for reliable clinical translation.

Conclusion

Korf’s account centers on making untargeted mass-spectrometry data more interpretable and reusable while emphasizing that computational tools cannot compensate for weaknesses elsewhere in the analytical workflow. The related clinical-metabolomics sources likewise describe standardization and quality practices as outstanding requirements for translation.