What Is a Mass Spectrometry Fragmentation Table?

A mass spectrometry fragmentation table is a structured record that pairs a parent molecule’s mass-to-charge ratio with the masses and relative intensities of the smaller pieces it breaks into when energy is applied inside a mass spectrometer. Think of it as a fingerprint card for a molecule: the parent ion sits at the top, and below it is a list of the fragment ions that result when specific chemical bonds snap. Researchers use these tables to identify unknown compounds, confirm the presence of known ones, and build the transition lists that power quantitative clinical and environmental assays. The concept sounds simple, but the details of how fragmentation tables are built, stored, and applied reach into nearly every corner of modern analytical chemistry.

How Molecules Break Apart Inside the Instrument

Mass spectrometers do not just weigh molecules. In tandem mass spectrometry, a selected precursor ion is isolated and then forced to collide with an inert gas like nitrogen or argon, a process called collision-induced dissociation (CID). The kinetic energy from those collisions is converted into internal energy, which causes specific bonds in the molecule to break. The resulting fragments are then sorted by their own mass-to-charge ratios and recorded as a spectrum. A fragmentation table is the tabulated version of that spectrum: each row lists a fragment’s mass, its relative intensity compared to the most abundant fragment, and sometimes the identity of the neutral piece that was lost.

The way a molecule falls apart is not random. Certain bonds are weaker and break preferentially. In peptides, for instance, amide bonds between amino acids are common cleavage sites, and the identity of the amino acids flanking those sites strongly influences which fragments form and how abundant they are. Amino acids like proline, tryptophan, aspartate, and arginine have an outsized effect on where amide bonds break, while residues like isoleucine, lysine, and histidine influence which fragments the detector actually sees even when they are not sitting right next to the cleavage point.¹ In small organic molecules, the story is different but equally predictable: functional groups like carbonyls, hydroxyl groups, and nitrogen-containing rings each have characteristic ways of fragmenting, and understanding those pathways lets analysts correlate a fragment pattern with a structural feature.²

What the Table Actually Contains

At its simplest, a fragmentation table is a two-column affair: fragment mass on the left, intensity on the right. But in practice, most useful tables carry more information. A precursor mass sits at the top, and the product ions beneath it are annotated with the neutral loss that produced them (for example, a loss of 18 mass units almost always means water left the molecule, while a loss of 28 often means carbon monoxide departed). In metabolomics workflows, each metabolite’s entry in a fragmentation table has been described as a precursor mass coupled with a binary string indicating which product ions are present and which are absent.³ That binary approach works well for screening large panels, where you only need to know whether a diagnostic fragment showed up, not exactly how intense it was.

Richer versions of the table record relative intensities as percentages, collision energy settings, ionization mode (positive or negative), and sometimes the instrument type. These details matter because the same molecule can produce noticeably different fragment patterns depending on the conditions used to break it apart.

Neutral Losses and Diagnostic Ions

Two concepts dominate how people read and use fragmentation tables. The first is the neutral loss: the uncharged piece that flies off the molecule and never reaches the detector. You infer its identity by subtracting the fragment’s mass from the precursor’s mass. Monitoring a specific neutral loss across many precursor ions is a powerful way to hunt for all molecules in a sample that share a particular functional group. This approach, known as neutral loss scanning, lets you selectively detect compounds bearing a given chemical feature within a complex mixture.⁴ If you scan for a loss of 176, for example, you are looking for glucuronide conjugates, a common metabolic modification in drug metabolism.

The second concept is the diagnostic ion, a fragment whose mass is so characteristic of a particular molecule class that its mere presence in a spectrum confirms what you are looking at. In lipid analysis, for instance, specific fragment masses correspond to individual fatty acid chains or headgroup fragments. Fragmentation tables for phospholipids record these diagnostic ions along with rule-based modifiers that describe how the fragment relates to the intact lipid’s structure. These rule sets form the foundation of computational tools that can generate predicted spectra for lipids that have never been measured experimentally.⁵

Why Collision Energy Changes Everything

The energy you pump into a molecule determines which bonds break. At low collision energy, only the weakest bonds give way, producing a few large fragments. Crank the energy up and those initial fragments break further, generating smaller and more numerous pieces. This means a fragmentation table built at one collision energy may look quite different from one built at another, even for the same compound on the same instrument.

Both the peptide’s sequence and the energy applied determine which fragment ions appear in the spectrum.⁶ Collision energy breakdown curves map out how each fragment ion’s intensity rises and falls as energy increases. Researchers use these curves to find the optimal energy for a given transition, since each precursor-to-product ion pair has its own sweet spot where signal is strongest.⁷ A new approach takes this even further by comparing how individual fragment intensities change across multiple collision energies, creating what amounts to a longitudinal profile for each fragment rather than a single snapshot. This multi-energy strategy has been shown to significantly improve the accuracy of identifying metabolites compared to matching spectra acquired at just one energy level.⁸

Spectral Libraries Are Collections of Fragmentation Tables

When thousands of fragmentation tables are gathered into a searchable database, the result is a spectral library. These libraries are the backbone of compound identification in fields from clinical toxicology to food safety. You acquire a fragmentation spectrum from an unknown compound, then search it against the library to find the best match. The growth of publicly accessible spectral libraries has been dramatic: the number of entries has increased more than 60-fold in roughly eight years.⁹

Building these libraries requires careful experimental work. A reference library for oligosaccharides in human milk, for example, was constructed using consensus spectra generated from multiple analyses, then made searchable through a specially designed matching algorithm.¹⁰ The idea of consensus spectra is important here: rather than relying on a single measurement, the library averages across many acquisitions to produce a more representative and reproducible reference pattern. This approach reduces the chance that an oddball measurement throws off future searches.

One major limitation of experimental libraries is that they can only contain compounds someone has already measured. The vast majority of small molecules in biological samples have no reference spectrum available. This gap has driven interest in computational prediction.

Predicting Fragmentation Tables Without Running Experiments

When a reference standard is not available, in silico tools try to predict what a molecule’s fragmentation spectrum would look like based on its chemical structure. One widely used approach, Competitive Fragmentation Modeling for Metabolite Identification (CFM-ID), is a machine learning tool that generates predicted tandem mass spectra for known or suspected metabolites without requiring a physical reference standard.¹¹ It relies on a statistical model trained on experimental spectra, so its accuracy is only as good as the diversity and quality of its training data.

Newer machine learning strategies aim to improve both speed and accuracy. A dual-model approach that combines a classification step for identifying real fragments and filtering noise with a regression step for predicting intensities has been shown to outperform earlier algorithms.¹² These tools are still far from perfect, especially for unusual molecular scaffolds that do not resemble anything in the training set, but they are increasingly useful for narrowing down candidates when searching a spectrum against a large chemical database.

Transition Lists for Targeted Quantification

Fragmentation tables have a very practical incarnation in clinical and pharmaceutical laboratories: the transition list. In targeted assays, you are not trying to identify unknowns. Instead, you already know exactly which compounds you want to measure, and you program the instrument to watch for specific precursor-to-fragment ion pairs, called transitions. Each transition is essentially one row from a fragmentation table: precursor mass on one side, product ion mass on the other, with the collision energy optimized for maximum signal.

This technique, known as selected reaction monitoring (SRM), uses the triple quadrupole mass spectrometer as a highly selective filter. The first quadrupole isolates the precursor, the second serves as a collision cell where fragmentation occurs, and the third passes only the chosen product ion to the detector. By spiking an isotopically labeled version of each target peptide into the sample, analysts can calculate the quantity of the native molecule by comparing signal intensities between the labeled and unlabeled versions.¹³ The choice of which transitions to monitor is critical: poor choices lead to interference from co-eluting compounds or weak signals.

Developing SRM assays used to be painfully slow because each protein required individual optimization. High-throughput methods based on synthetic peptide libraries have made it possible to generate transition lists for hundreds of proteins at once.¹⁴ These pre-built transition tables are shared across laboratories, enabling reproducible measurements of the same proteins in different clinical settings.

Different Ways of Breaking Molecules Produce Different Tables

CID is the most common fragmentation method, but it is not the only one. Electron transfer dissociation (ETD) and electron capture dissociation (ECD) break bonds through fundamentally different mechanisms, and the resulting fragment patterns look quite different from CID spectra. A comprehensive comparison of CID and ETD on roughly 19,000 peptides found that CID identified about 50 percent more peptides overall, but ETD produced roughly 20 percent more sequence coverage per peptide. Combining both methods pushed average sequence coverage to 92 percent.¹⁵

ECD, available on specialized instruments, can yield even higher fragment counts and spatial resolution than either CID or ETD, making it valuable for applications like oxidative protein footprinting where pinpointing exactly which residue was modified matters.¹⁶ The practical implication is that a fragmentation table is not universal for a given molecule; it is specific to the fragmentation method used. A CID table and an ETD table for the same peptide will share some entries but differ substantially in others.

Hybrid strategies that combine spectra from multiple fragmentation methods into a single composite spectrum are gaining traction, particularly for glycoproteins. One such approach concatenates ETD and higher-energy CID spectra acquired back-to-back on the same precursor ion, capturing the best identification features from both methods while preserving reporter ion quality for quantification.¹⁷

Using Fragmentation Predictions to Improve Database Searches

In proteomics, identifying a peptide usually involves comparing an experimental spectrum against a database of theoretical spectra generated by applying simple fragmentation rules to every possible peptide sequence. The problem is that these theoretical spectra are crude: they assume every fragment ion is equally likely, which is far from reality. Hidden Markov models and other statistical tools trained on large experimental datasets can predict which fragment ions are likely to be intense for a given peptide sequence. Rather than using these predictions as literal replacement spectra, researchers have found that the real value lies in using them as a filter, flagging fragment ions that are unlikely to appear so the search algorithm can ignore them and focus on the ones that matter.¹⁸ This pre-filtering step improves the statistical power of the match between an experimental spectrum and the correct peptide candidate.

The Isomer Problem

Fragmentation tables work beautifully when two molecules differ enough in structure that their fragments are distinct. They struggle when molecules are structural isomers, meaning they have the same molecular formula but different arrangements of atoms. Positional isomers sometimes produce unique fragments that a table can capture, but stereoisomers (mirror-image molecules or molecules differing only in the spatial orientation of a group) are a harder problem. For chiral stereoisomers, unique fragment masses simply do not exist regardless of which bonds are broken, so any differences between their spectra can only come from variations in fragment intensities rather than the presence or absence of specific fragments.¹⁹ Distinguishing stereoisomers therefore requires statistical methods sensitive enough to detect subtle intensity shifts, and even then, the differences can be smaller than run-to-run measurement noise.

This limitation explains why mass spectrometry is sometimes paired with chromatographic separation or ion mobility measurements. A fragmentation table alone cannot always tell you which isomer you have, but combined with retention time or collisional cross-section data, the identification becomes much more reliable.

Standardized Formats for Sharing Tables

A fragmentation table is only as useful as your ability to share it with other laboratories and software tools. For years, spectral libraries were locked in proprietary vendor formats, making it difficult to compare data across platforms. The Proteomics Standards Initiative has developed standardized formats to address this, including mzSpecLib for encoding spectral libraries and mzPAF for annotating fragment ion peaks. These formats use a flexible, extensible data model that can be converted between different file types, making it easier for researchers to exchange fragmentation data across instruments and software ecosystems.²⁰

Standardization also helps with reproducibility. When two labs use different annotation conventions for the same fragment, automated comparison fails even though the underlying data agree. Having a controlled vocabulary for peak labels and metadata fields means a fragmentation table generated in Tokyo can be searched against a library curated in Boston without manual reformatting. For a field that increasingly depends on large-scale data sharing and machine learning, these format standards are quietly foundational, even if they rarely make headlines.