NEW YORK – Due to the SARS-CoV-2 pandemic, this year's American Society for Mass Spectrometry annual meeting was a virtual one, with vendors making their traditional product introductions via remote events held during the past two weeks.
The meeting was also a somewhat quiet one for new product introductions, with much of the focus on extending the capabilities of existing platforms through new workflows and software.
One major point of emphasis was the incorporation of machine learning and deep learning into the peptide identification process. For years, researchers have used machine learning-based software tools like Percolator to improve the confidence of peptide identifications in their experiments, and in recent years such efforts have accelerated with a number of research groups presenting deep-learning tools for predicting patterns of ion fragmentation, which could likewise help improve the confidence of peptide identifications and allow researchers doing data-independent acquisition mass spec to run such experiments without first generating sample-specific spectral libraries.
During its ASMS presentation, Thermo Fisher highlighted the incorporation of one of these tools — deep-learning software from German proteomics informatics company MSAID based on the Prosit package developed by researchers at the Technical University of Munich — into its new Proteome Discoverer 2.5 software package.
In a typical proteomics experiment, peptides are fragmented to produce a set of fragment ions from which mass spectra are generated. These experimental spectra are then matched to a database of theoretical spectra, allowing researchers to identify the peptides and, ultimately, proteins in a sample.
Interpreting these spectra requires an understanding of how particular peptides fragment, and while researchers have a general understanding and ability to predict this process, predicting what ions will be produced at what levels of intensity remains a challenge. As such, many software tools for matching peptide spectra assume that all possible peptide ions are equally likely to be produced and at the same intensity.
Deep-learning approaches offer the possibility of improved peptide-spectra matching by allowing researchers to train software to better understand the specific fragmentation patterns of specific peptides under particular conditions.
Bruker also highlighted during its conference presentation work using deep learning to improve peptide identifications. Unlike the approach taken by Thermo Fisher with its Proteome Discoverer software, Bruker is using ion mobility data generated by its timsTOF Pro mass spec, to provide an additional dimension of information that can be used for identifying molecules.
The behavior of peptides in an ion mobility system can be described by use of their collisional cross sections (CCS), with each peptide having unique collisional cross-sectional profiles. Bruker and other vendors, like Waters, who have made prominent use of ion mobility technology, have been exploring the use of CCS data for helping with peptide identification.
At ASMS, the company presented work by the lab of Max Planck Institute of Biochemistry professor Matthias Mann, in which researchers collected 2 million CCS values from 360 mass spec runs across cell digests from five organisms and used that data to train a neural network to predict CCS values based on peptide sequence.
"If you know what the tryptic peptides are, you can predict the collisional cross sections," said Rohan Thakur, executive vice president of life science mass spectrometry at Bruker.
"Fundamentally, [the Mann lab] is using machine learning to amplify what the technology can deliver," he said. "You need confidence in [identifications] where you have poor signal to noise. Everything works when you have strong signal to noise. But most small changes often occur in the noise, [which] is where you want additional parameters to improve your confidence. If we can predict these soft spots, predict phosphorylations, predict glycosylations, and then go look for them in the acquired data, it improves your depth of coverage."
"I think it’s a general trend in the industry," Andreas Huhmer, director of omics at Thermo Fisher Scientific, said of the incorporation of machine- and deep-learning tools. "Having a tool available that can predict exactly what a spectra should look like for a particular sequence, you can look through your data very quickly [to make IDs]. And that improves the confidence of the data."
He said that for normal proteomic experiments, the company sees around a 10 percent to 20 percent improvement in the number of peptide identifications using the deep-learning approach.
Perhaps more significant than this boost in identifications, though, is the potential of such approaches to enable new kinds of experiments with peptide search spaces that were too complex for previous methods, Huhmer said, citing examples like immunopeptidomics and metaproteomics as fields he expected the emerging deep-learning approaches to play a role in advancing.
Iain Mylchreest, VP of R&D, analytical instruments at Thermo Fisher, noted that the company was also investing in more sophisticated approaches to acquiring mass spec data. As an example of this, the company said during its ASMS presentation that it was including its AcquireX workflow into its new Xcalibur 4.4 software package, making it available on its Orbitrap Exploris platforms. The workflow allows researchers to select from run to run which molecules the mass spec should and shouldn't focus on, allowing for deeper analysis of relevant analytes, Mylchreest said.
Bruker likewise introduced new mass spec acquisition approaches, including a real-time search engine based on the ProLuCID database search tool developed by Scripps Research Institute Professor John Yates.
The tool will allow for more intelligent searching, Thakur said. "You searched, you saw this peptide, you identified the peptide, you know it has serine, threonine, and tyrosine, which are classic hotspots for phosphorylation. Now you can structure your acquisition to go look for these post-translational modifications.
"If you have intelligence built into your acquisition software, you don't just depend on what the machine sees, but you become more predictive," he said. "You can look ahead and target precursors more intelligently rather than doing then randomly based just on their intensity in the mass spec."
New Instrumentation Releases
While new mass spec instrumentation was not a major theme of the meeting, there were a few releases. Most notably, Thermo Fisher launched its Orbitrap Exploris 120 and 240 mass spectrometers, both of which are intended mainly for routine analyses.
The company launched the Exploris line at the 2019 ASMS meeting with the release of the Exploris 480, a higher-end research instrument, which it has positioned as an extension of its Q Exactive line.
The Orbitrap 240, Huhmer said, is suitable for proteomics work but is intended to handle more routine protein profiling experiments.
"The 480 has a lot more sensitivity, so if you are looking at rare samples even down to a single-cell sample, the 480 will have more sensitivity and give you the better results," he said. "But this particular instrument is squarely focused on making sure core labs have all the functionality and at a very high performance level that they typically need."
Agilent also highlighted a new mass spec platform, its Agilent Infinity II Clinical Edition K6460S, a system it launched this spring as a class I device in the US and a class II device in China. Aimed at laboratory-developed test development, the release marks a step in the company's efforts to streamline clinical mass spec, joining the broader industry trend.
Agilent also figured into the ASMS presentation of chromatography firm Mobilion, which said that it would bring its structure for lossless ion manipulation (SLIM) ion mobility separations technology to market in partnership with Agilent. The company plans to launch the SLIM system on Agilent's QTOF mass spectrometers next year.
Developed at Pacific Northwest National Laboratory, the SLIM technology extends ion mobility path lengths beyond those allowed by conventional ion mobility spectrometry systems, potentially enabling much more extensive separations, which could aid in a range of mass spec-based research areas, including characterization of biologic drugs and protein biomarker discovery.
Melissa Sherman, CEO of Chadds Ford, Pennsylvania-based Mobilion said during the ASMS presentation that the company was focused on three main areas — biopharma, liquid biopsy, and early disease detection.
"What [customers in those areas] tell us is that they often experience a tradeoff," she said. "They might get resolution or throughput or ease of use, but there really isn't one instrument platform that provides all [three] in one. If they want higher resolution, it's often a tradeoff with throughput. If they want something that is very easy to use, typically the performance of those systems doesn't provide the resolution they are looking for. We're setting out to… eliminate those tradeoffs."
The device is conceptually similar to Waters' Select Series Cyclic IMS instrument, which uses an IMS device with a circular path that allows researchers to cycle ions of interest through multiple times to achieve higher resolution separations. Waters introduced the device at last year's ASMS meeting. The company did not announce any new product launches at this year's meeting.
Sciex did not launch a new mass spec instrument, but the company did announce the official launch of its Echo MS System, which uses acoustical liquid handling to introduce up to three samples per second to a mass spectrometer.
Developed in collaboration with Sciex's fellow Danaher firm Labcyte, the system is intended for large-scale screening experiments in areas like drug development and biomarker research and can work with a variety of analytes including peptides.