Enhancing Mass Spectrometry Data Accuracy

I got my PhD in electrical chemistry from the University of Washington. Then I went to Perkin Elmer for 10 years to do R&D, focusing on computers, data processing algorithms, software and doing optical spectroscopy. It had nothing to do with mass spec. I didn't know mass spec at all.

My introduction to mass spec came unexpectedly. Management asked the R&D team to tackle a major production issue in the GC-MS factory. We were put under intense pressure to find a solution and in the end, we were able to fix the problem and save the group. Through that experience, I became interested in mass spec.

Those of us coming from an optical spectroscopy background were surprised to see how heavily mass spec data was processed into centroid form before analysis. While centroiding reduces data size, it also strips away much of the signal detail, making it harder to distinguish true peaks from noise. We saw an opportunity to improve the data quality and accuracy and along the way invented a few patents.

Contrary to general-purpose software companies that focus on building faster workflows or stitching together different tools, we pride ourselves in being able to actually improve the accuracy of the data in the process, not just simply processing it as is.

How do we do that? Basically, it’s a new way to calibrate mass spec systems to significantly enhance their accuracy. It turned out we could improve the accuracy by a factor of 100 on lower resolution systems. This means you can get performance that rivals high-resolution instruments without the hefty price tag. Instead of buying new equipment, labs can extract more performance from the systems they already have.

What we found is that, for historical reasons, mass spec data often undergoes heavy onboard processing. By the time the user sees the data, it can already be somewhat corrupted, though most users don’t even realize it. So, we went back to fundamentals.

Our process starts with raw profile data, before it’s been heavily processed. We developed a new way to calibrate this raw data. All vendors calibrate mass spectrometers, but they typically just shift the mass axis, the x-axis. We go a step further and calibrate both the x-axis and the y-axis, so the entire spectral curve becomes accurate. Once the full spectral curve is more accurate, you can start to resolve subtle spectral differences, even on lower-resolution instruments, without needing to invest in high-resolution systems.

Exactly. There’s a calibration stage, and then there's the data processing that happens directly on the raw data. That algorithm gives you much tighter control over the final result.

That's why you can discern small spectral differences between very similar compounds. If you’re trying to identify two compounds that are structurally close, now you can separate them. You can actually tell them apart. And by the same token, if you’re dealing with a mixture where components aren’t well separated due to interference, our approach lets you account for those mixtures. We've developed a new way to handle mixture analysis. 

It's especially powerful in cases where you're dealing with unexpected or unknown components. If it's a routine, very tightly controlled process, we might improve the front end and clean up the data a bit, but there's no new information insight. Our strength really shows when there's a process control issue or you're dealing with the unknown.

We actually offer two main solutions. One is almost like a teaching tool. It gives you such tight resolution that when things are closely overlapping, you can visually work through the data and start to understand the chemical associations; how the fragments relate, what the signatures mean. It’s something that’s best seen in action, but it leads to those “aha” moments where you suddenly grasp why certain compounds are likely neighbors.

Then we have customers who need it to be fully automated. You just run the algorithm overnight, and by morning you’ve got a complete batch report. It’s matched against the spectral library, down to the wire level. So, we support both: one that’s interactive and educational, empowering scientists, and one that’s push-button automation.

First of all, it’s very easy to see if you have spectral interference because our software will show it. If there’s interference, that means there’s an extra signal, and when that happens, you’ll see a drop in spectral accuracy. You can even visualize this interactively in the software. You’ll be able to see where the interference is coming from, and in many cases, postulate what it might be. This lets you identify not just your main components, but also any impurities or interferences, whether or not they’re related to your target compound. That’s the power of spectral accuracy. First, it alerts you that interference is present. Then it helps you dig deeper, potentially identifying what that interference is.

For example, losing a hydrogen in an electron ionization source is a very common source of interference. It’s not coming from the chromatography or from user handling, it happens in the ion source itself. The molecule gets hit with electrons and can lose a hydrogen atom. That 'minus H' species can spectrally overlap with your main peak, sometimes as much as 50/50 or more, which is remarkable. Our software is uniquely equipped to handle these kinds of challenges. It can tell you whether that type of interference is present, where it occurs, and what it’s likely to be.

Cerno’s customer base is remarkably diverse, with more than 2,000 software licenses in use across the global market today. While our solutions are widely adopted in academic research settings, a significant portion of our user base comes from industry and government labs.

Major pharmaceutical companies such as AstraZeneca, Pfizer, Lilly and Novartis rely on Cerno’s software to enhance their mass spectrometry workflows. We’re also proud to support regulatory and public health agencies, including the US Environmental Protection Agency and the California Department of Public Health.

On the academic side, our tools are used at top-tier institutions including Stanford University, Scripps Research, UC Davis and the University of Zurich, among many others. This broad adoption reflects our platform’s flexibility and value across discovery, analytical development, forensics, and applied research in pharma, biotech, government, and academia.

We’re really looking at growth in two dimensions. First, there’s the software experience itself. We’re exploring ways to make it more accessible, like through a cloud-based environment. That would let users check in and check out easily, with flexible licensing options like site or company-wide licenses. We don’t offer that today, but we know it would enable people to share their experience more dynamically and even work from home.

The second dimension is international expansion. We’re looking to build relationships with the right partners, those who can educate users, facilitate adoption, and remain vendor-agnostic. We want partners who are present in the labs, who can provide service, support, calibration and training in a neutral way. It’s not easy to find partners who have both the technical expertise and the trusted advisor status we need, but that’s a key focus for us moving forward.

Yes, absolutely. Our technology is not only compatible with existing instruments but is also being integrated into new platforms. A prime example is the recently launched InfinityLab LC/MSD iQ+ from Agilent, which comes with our MassWorks® software built in – making it the first mass detector to offer native, accurate mass calibration without the need for high-end hardware.

Beyond that, our software has been successfully used with both LC-MS and GC-MS instruments from virtually every major mass spectrometry OEM in the market. This broad compatibility reflects the versatility and value of our platform across a wide range of applications and instrument types.