A Swiss-based hospital system has adopted new software from Geneva Bioinformatics in combination with liquid chromatography tandem mass spectrometry, or LC-MS/MS, to conduct emergency clinical toxicology screening.
Last month, GeneBio announced that the clinical toxicology laboratory at Geneva University Hospitals, better known by the acronym HUG, will use the company's newly launched SmileMS spectral library search software for LC-MS data to quickly identify toxic compounds in patient samples in the emergency room setting.
The software is based on X-Rank, an algorithm developed by researchers at the Swiss Institute of Bioinformatics, HUG, and GeneBio. SmileMS lets users identify small molecules with fewer false positives than other mass spectra analysis tools, according to the company. It also screens positive and negative ions in a single run, searches multiple spectral libraries, and generates a report in a single format. Another differentiator the company highlighted is the fact that the software can analyze spectra produced by different MS instruments.
GeneBio noted that most algorithms that are currently used to identify small molecules via MS/MS spectral library searches were developed for gas chromatography mass spectrometry, not LC-MS/MS, which is gaining ground among research labs. SmileMS, by comparison, was developed specifically for analyzing LC-MS data.
In terms of toxicology screening, the company said that most clinics currently rely on liquid chromatography with ultraviolet diode array detection, as well as methods like immunoassays and GC-MS, but LC-MS/MS-based testing combined with spectral library searching has "proven itself to be a quick and powerful diagnostic technique," providing "unambiguous compound identification" within 20 minutes.
Marc Fathi, the head of the toxicology lab at HUG, told BioInform that LC-MS/MS is a useful technique for small-molecule identification because it is a "universal method" that can detect molecules of different sizes. He said that originally, LC-MS/MS was used mostly in specialized research laboratories, primarily because of the complexity of the method and the instruments, but that in the last five to six years, routine laboratories have begun to use the method more often.
Pierre-Alain Binz, senior scientist and SmileMS product manager at GeneBio, told BioInform that having access to a larger range of molecules that weren't available for GC-MS-based applications formed "the starting material" for developing SmileMS and that the developers "had a clear challenge to develop an algorithm that is efficient enough to deal with the heterogeneity of the LC-MS/MS data."
A third challenge for the developers, Binz said, was to develop software that can be used by routine lab technicians as well as trained chemists.
Olivier Philippe, GeneBio's sales and marketing director, concurred, adding that the developers were also tasked with the responsibility of delivering a vendor-neutral software that could handle spectra from different types of LC-MS/MS instruments.
Currently, mass spec vendors like Thermo Fisher Scientific, Waters, Agilent Technologies, and AB Sciex offer pre-packaged software for handling the spectra their instruments produce, as well as spectral libraries for comparison.
However, there is "no product like [SmileMS] on the market that is vendor neutral," Binz said. "Vendors have their own solution … but there is no other generic solution on the market."
Philippe added that many labs have more than one instrument type. "If you want to search a library with [spectra from] different instruments, then SmileMS is very appropriate for this." He said that the software includes an "extensibility feature" that makes it easier for users to create their own libraries of compounds using data from multiple instruments.
In addition to its work with HUG, the company is also working with Hans Maurer, head of the department of experimental and clinical toxicology at Germany's Saarland University, to develop an LC-MS library of drugs and poisons for use in toxicology testing.
Maurer is the co-developer of the standard in the field, the Mass Spectral Library of Drugs, Poisons, Pesticides, Pollutants, and Their Metabolites, which is based on GC-MS spectra. He has been testing SmileMS against the new library for the last year and noted in a company statement that it "shows good performances for targeted and untargeted identification."
Not a Dot Product
Binz explained that most software used to search and compare mass spectra, such as ThermoFisher's ToxID, AB Sciex's Cliquid, and NIST's MS Search, are based on a dot-product algorithm that relies on the intensities of MS signals to score a spectral match.
"Difficulties occur when an MS signal is predominant in a spectrum," he said. "Dot-product-based software tend to give too much weight to the matching and [don’t discriminate] between different molecules sharing the same signal. In that case minor peaks are underestimated."
In a paper published last September in Analytical Chemistry describing the X-Rank algorithm that underlies SmileMS, Binz and his colleagues wrote that X-Rank "first sorts peak intensities of a spectrum and second establishes a correlation between the two sorted spectra. [It] then computes that probability that a rank from an experimental spectrum matches a rank from a reference library spectrum."
X-Rank uses "intensity ranks and applies a log-likelihood ratio approach to score a match" and "calculates a ratio between a probability of a correct match and the probability of a random match" Binz explained. Software tools that use the dot-product approach, he said, "have no notion of probability."
In the paper, the authors note that while dot-product algorithms "perform well in certain applications," they "do not properly address the questions associated with LC-MS/MS."
For example, MS Search "is well adapted to GC/MS data but cannot cope with the high
variability of LC-MS/MS." Cliquid, meantime, is "mainly run with [AB Sciex] instruments and proprietary data, which are acquired with the same instrument," making the software "not transferable across platforms," the authors said.
In a set of tests described in the paper, the researchers compared X-Rank to MS Search using a dataset containing 1,050 product ion spectra from more than 300 compounds generated with the AB Sciex QTRAP, and another dataset containing 4,504 product ion spectra from more than 1,000 different substances generated from a Paul Trap machine.
The two data sets were used as both input and reference, resulting in a total of four comparisons. In all four cases, "X-Rank showed better sensitivity and specificity" than MS Search, especially for cross-platform identification, the researchers wrote in the paper. For example, in one test, X-Rank correctly identified 3,932 molecules while the MS search identified 3,916.
Binz said users can run the web-based SmileMS software locally or access it remotely using a graphical user interface that has been tested and evaluated by several beta testers. This makes it possible, he said, for lab technicians to use the software to collect results and generate reports for clinicians and for more expert users to "evaluate the results in terms of what compounds were identified [and] with what level of confidence … and to browse the actual libraries that were searched." He noted that users with more expert training can also annotate and modify existing libraries.
Fathi said that researchers at HUG will use the software to identify drug molecules in patient samples. "For the identification we have constructed a library of substances and the comparison tool to [make the] identification is SmileMS," he said. "The software [compares] data obtained from a sample that was investigated for possible identification and the library of spectra we have on the instrument."
To develop the library used to compare spectra in the hospital, Fathi's team analyzed and generated mass spectra for 160 compounds and compared their findings to spectra published in the scientific literature. He said that the team will continue to add additional compounds to the library and that it will be used only by HUG's researchers and not marketed with the software.
Philippe said that the primary market for SmileMS is researchers in forensic and clinical toxicology as well as researchers who work in food control, doping, metabolomics, and more general pharmaceutical applications.
He added that the company is currently seeking "alliances with MS vendors to promote and possibly distribute SmileMS with their software platforms."
Neither Binz nor Philippe would provide additional details about the terms of the agreement with HUG or on pricing for the software.