For many analytical laboratories, the archive is both a treasure chest and a graveyard. Years of LC–MS and GC–MS data sit untouched, locked behind outdated software, incompatible formats, and forgotten acquisition parameters. “In theory, you have a gold mine from an old study,” Ansgar Korf, CEO and Co‑Founder of mzio GmbH, remarks, “but if the tech stack doesn’t work anymore, then the value is gone.”
Korf and his team at mzio are working to change that. Building on the long‑established mzmine platform, they are developing tools and frameworks that make mass spectrometry data FAIR—findable, accessible, interoperable, and reusable—at enterprise scale. Their new initiative, FAIR-MS, aims to help labs finally unlock the scientific value buried in their historical datasets.
Why Labs Struggle to Reuse Old MS Data
Korf is direct about the core challenge: mass spectrometry data is inherently complex, and the tools to interpret it have not kept pace with the volume being generated.
“Looking at mass spec data is quite a task,” he notes. “If you do it manually, you get stuck pretty quickly.” Even when software exists, labs often face a second barrier: simply accessing the data. Pulling a five‑year‑old dataset from an archive can feel like digital archaeology. “You still don’t know how that data was acquired or whether you even have software that can read it.”
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
Vendor lock‑in and shifting file formats only deepen the problem. While open formats such as mzML help, they still require conversion steps that slow down workflows. mzio’s approach is to read native vendor files directly, eliminating the conversion bottleneck and making historical data immediately usable.
What FAIR Means in Daily Lab Practice
FAIR principles are widely discussed in academic circles, but Korf emphasizes their practical value for working scientists. “FAIR principles make data useful,” he asserts. “If you cannot access the data by any blocker, then it’s worthless.” In daily practice, FAIR means that a lab should be able to retrieve, read, and analyze any dataset, regardless of age, instrument vendor, or software version.
Without that, labs waste time and money reacquiring data they already possess. Korf warns that this is really a waste, especially when older datasets may contain insights that were not detectable with earlier algorithms.
The Challenge of Vendor‑Neutral Workflows
Mixed‑vendor labs are now the norm, not the exception. But building workflows that operate seamlessly across different instrument ecosystems remains technically demanding.
According to Korf, the hardest barriers fall into two categories:
- Different data types require different algorithms: “Ion mobility data needs to be treated differently than GC‑EI data,” he explains. A single workflow must accommodate multiple separation dimensions and acquisition modes.
- Raw data accessibility: Even with open formats, conversion steps slow down analysis. “Having a tool that can read the native data files reduces another processing step,” he notes. “It speeds things up again.”
This is where mzio sees a major opportunity: providing a unified software layer that abstracts away vendor differences and lets scientists focus on the chemistry, not the file formats.
Calibration, Standardization, and Confidence in Multi‑Dimensional MS
As mass spectrometry becomes increasingly multidimensional (combining retention time, m/z, ion mobility, CCS, imaging, and fragmentation), the need for accurate calibration grows.
Korf highlights mzio’s collaboration with the Center for Mass Spectrometry and Optical Spectroscopy (CeMOS) in Mannheim, Germany, and Polymer Factory in Stockholm, Sweden, where they are developing integrated calibration mixtures for lipidomics. He notes that oxidized lipids, in particular, present a challenge because everything looks very similar in terms of fragmentation and MS¹ data.
The goal is to unify calibration across mass accuracy, retention time, and ion mobility using a single, software‑integrated standard. “The more accurate you are,” Korf emphasizes, “the higher the chance you’ll actually find something useful.”
What AI‑Ready Mass Spectrometry Data Really Requires
AI is the topic everyone wants to discuss—but Korf is careful to separate hype from practical reality.
In close collaboration with academic partners, mzio has gained expertise in training and using deep learning models based on data from public MS repositories, but the quality of public data is inconsistent. FAIR-MS aims to solve this by enabling companies to extract the most from their own internal data to create a high quality foundation for proprietary AI models.
This matters because AI models trained on metabolomics or natural products may not generalize to pharmaceutical or chemical industry datasets. “We want a model created inside one enterprise with their data,” Korf explains, “where we can control the quality and train models based on their actual needs.”
AI also plays a role in usability. mzio uses vector databases and embedding‑based search to make workflows faster and more intuitive. “This would simply not work without AI,” he adds.
Shifting the Focus of the Field
Mass spectrometry has long been driven by hardware innovation—higher sensitivity, faster acquisition, and more dimensions. Korf believes it’s time for software to shine.
“Traditionally the hardware is always in the spotlight,” he observes. “But now the amount of data is becoming more vast. It’s time to really extract the most out of it.”
The future, he argues, lies in tight integration between instruments and intelligent software that can interpret the data deluge.
Conclusion
Korf’s message is clear: the mass spectrometry community has invested heavily in generating data, but far less in making that data usable. FAIR principles, vendor‑neutral workflows, and AI‑ready data structures are no longer academic ideals—they are practical necessities.
As instruments continue to evolve, the real breakthroughs may come not from the hardware, but from the software that finally unlocks the full value of the data we already have.
