Fragment Coverage: Unified Visualization and Quantitative Assessment of MS/MS Fragmentation Supports Metabolite Identification in Macromolecule Drug Metabolism Studies Authors: Paula Cifuentes, Albert Garriga, Fabien Fontaine, Ismael Zamora Published: 2026 ABSTRACT ABSTRACTRationalePeptides and oligonucleotides are increasingly used as therapeutic agents, making the identification and structural elucidation of their metabolites a critical step in drug development. Liquid chromatography–mass spectrometry (LC‐MS/MS) is the primary analytical technique for this purpose; however, extensive backbone fragmentation generates highly complex MS/MS spectra, making spectral interpretation challenging and limiting the exploitation of the structural information contained in MS/MS fragmentation data.MethodsThis study introduces two fragment‐based metrics, fragment coverage and complete fragment coverage, to quantify structural coverage derived from MS/MS fragmentation at metabolite and parent compound levels. Their performance was evaluated across 33 LC‐MS/MS experiments comprising peptide and oligonucleotide datasets. Ranking analysis based on mean average precision (MAP) assessed discrimination between true and false metabolite assignments. Additionally, a fragment viewer was developed to map MS/MS fragment ions onto structures, facilitating visualization and spectral interpretation.ResultsA total of 255 metabolites were identified, including 159 true metabolites and 96 false positives. True metabolites consistently exhibited higher fragment coverage and complete fragment coverage values than false positives. This trend was supported by ranking analysis using mean average precision (MAP), where complete fragment coverage achieved the highest performance (0.9737), followed by fragment coverage (0.9676), MassMetaSite score (MMS score) (0.9390), and isotopic similarity (0.9090). Two representative case studies demonstrated the applicability of the proposed metrics.ConclusionsThe proposed fragment coverage metrics enable quantitative assessment of MS/MS fragmentation, providing complementary information to support manual interpretation of LC‐MS/MS data in MetID workflows. Together with the fragment viewer tool, these approaches facilitate more efficient and interpretable analysis of complex MS/MS spectra in macromolecules.Keywords: LC‐MS/MS, macromolecule metabolite identification, MS/MS fragmentation, oligonucleotides, peptides Introduction In recent years, large therapeutic biomolecules such as peptides and oligonucleotides gained increasing relevance as therapeutic agents [1, 2]. A critical step in their development is the identification and structural characterization of metabolites and degradation products, which is essential for evaluating stability, pharmacokinetics, safety, and therapeutic potential [3, 4]. LC‐MS/MS is one of the primary analytical techniques for metabolite identification (MetID) in the pharmaceutical industry, as it combines high sensitivity, speed, and compatibility with automation workflows [5]. In MetID studies, structural information about the identified metabolites is obtained from MS/MS spectra, generated by isolating a specific molecular ion and inducing controlled fragmentation [6]. This information is interpreted through characteristic fragment ion series that depend on the precursor ion and its charge localization. Backbone fragmentation in peptides and oligonucleotides generates characteristic fragment ion series. In peptides, fragmentation along the backbone produces a–c ions that typically retain the charge on the N‐terminal fragment, whereas x–z ions retain it on the C‐terminal fragment; oligonucleotide backbone fragmentation generates a–d ions derived from the 5′ terminus, and w–z ions derived from the 3′ terminus [7, 8]. All these ions can be generated in different charge states, leading to highly complex MS/MS spectra with overlapping peaks and variable ion intensities. Consequently, interpretation of MS/MS fragmentation often involves a combination of automated annotation tools and manual peak‐by‐peak inspection of spectra to determine which regions of the parent and metabolite structures are supported by fragment ions. Due to this spectral complexity, the process is both challenging and time‐consuming [9]. As a result, this complexity limits the practical exploitation of MS/MS fragmentation data in MetID workflows, despite its potential to provide detailed structural information that can complement other analytical criteria for distinguishing true from false metabolite assignments. Coverage metrics have long been used in proteomics to quantify experimental evidence supporting molecular identification. In bottom‐up proteomics, protein sequence coverage is defined as the proportion of a protein's amino acid sequence that is represented by the set of experimentally identified peptides relative to the full reference sequence. This metric reflects the extent to which a known protein sequence is experimentally supported by peptide level evidence [10]. At the MS/MS level, peptide identifications are supported by characteristic backbone fragmentation patterns. The extent to which an observed spectrum can be explained by fragment ions is commonly quantified as the proportion of theoretical fragment ions or backbone cleavages that are observed in the spectrum, reported in some studies as peptide sequence coverage [11]. Similarly, in top‐down proteomics, sequence coverage is expressed as the fraction of backbone cleavage sites supported by observed fragment ions, reflecting the extent to which the intact protein sequence is supported by experimental fragmentation evidence [12]. Although these coverage metrics provide valuable measures of experimental support for protein and peptide identification, they are not directly applicable to MetID. In contrast, MetID of therapeutic compounds focuses on the structural elucidation of biotransformation products from a known parent compound. The resulting metabolites often exhibit chemically diverse and partially unpredictable modifications, leading to multiple plausible structural candidates. Consequently, although coverage metrics have proven valuable in proteomics, analogous quantitative measures for evaluating structural support during therapeutic peptide and oligonucleotide MetID remain lacking. To address this need, this study introduces two fragment‐based metrics, fragment coverage, and complete fragment coverage, designed to quantify structural coverage derived from MS/MS fragmentation in peptide and oligonucleotide MetID at both metabolite and parent compound levels. Fragment coverage is conceptually analogous to sequence coverage in top‐down proteomics and to protein sequence coverage in bottom‐up proteomics, extending these ideas to metabolite structural elucidation rather than protein identification. Complete fragment coverage quantifies how extensively the observed MS/MS spectrum represents all theoretically possible fragment ions of the molecule. The proposed metrics are evaluated across 33 LC‐MS/MS experiments comprising peptide and oligonucleotide datasets. Their ability to support manual data interpretation is assessed based on their capacity to discriminate between true and false metabolite assignments using ranking analysis. In this evaluation, two established MetID properties [13], routinely used as quantitative criteria in expert data interpretation, are included as reference alongside the proposed metrics. Finally, to facilitate the interpretation of MS/MS spectra, a visualization tool referred to as the fragment viewer is also introduced, which maps all the assigned MS/MS fragment ions directly onto parent and metabolite structures. Methods To provide the necessary context for the fragment‐based metrics introduced in this study, we briefly describe the structure elucidation and fragment assignment strategy used in the processing of raw LC‐MS/MS data. The resulting annotated MS/MS fragment ions serve as the basis for the calculation of the two proposed metrics, as well as for the visualization approach. Assignment of MS/MS Fragment Ions Structure Elucidation and Fragment Assignment In our previous study, we presented an automated LC‐MS/MS data analysis workflow specifically designed to address the challenges of MetID for macromolecules [13]. MassMetaSite (MMS) helped to create a faster system for the data processing step. It selects the chromatographic peaks that are related to the compound of interest. It then finds the mass spectral information for each extracted peak. Next, it assigns potential structures by comparing theoretical fragmentation (predicted from the structure and fragmentation of the parent molecule) with the experimental mass‐to‐charge ratio (m/z) values obtained from the spectra. Finally, it scores potential solutions based on the fragments assigned to the spectra alone or by comparison with the parent fragmentation [14]. A full description of the raw LC‐MS/MS data processing is provided in the reference study [13]. Here, we briefly summarize the fragment assignment and structure elucidation, which is performed using two complementary strategies: Parent‐based fragments: The parent molecules are virtually fragmented at each intermonomer bond following the defined ion series for peptides (a–c and x–z) and oligonucleotides (a–d and x–w), generating theoretical m/z values. These theoretical fragments are then compared with the parent spectra, and any peaks in the parent spectra that match the predicted m/z values within a certain mass tolerance are selected for the next step in the methodology. The structural fragments identified in the parent spectra are mapped onto the metabolite structure to determine whether they correspond to a shifted, nonshifted, or partially shifted fragments. Once the metabolite fragment structures are defined and their theoretical m/z values computed, these are compared to the metabolite spectra to identify substrate‐metabolite fragment pairs. Peaks are classified as matches (red) if the observed m/z corresponds to expected fragments (shifted or unshifted due to metabolism) or as mismatches (blue) if the observed m/z is unshifted but the fragment contains a chemical modification that would alter the m/z. Peaks not used for structure assignment are displayed in black [15]. Parent‐based fragments: The parent molecules are virtually fragmented at each intermonomer bond following the defined ion series for peptides (a–c and x–z) and oligonucleotides (a–d and x–w), generating theoretical m/z values. These theoretical fragments are then compared with the parent spectra, and any peaks in the parent spectra that match the predicted m/z values within a certain mass tolerance are selected for the next step in the methodology. The structural fragments identified in the parent spectra are mapped onto the metabolite structure to determine whether they correspond to a shifted, nonshifted, or partially shifted fragments. Once the metabolite fragment structures are defined and their theoretical m/z values computed, these are compared to the metabolite spectra to identify substrate‐metabolite fragment pairs. Peaks are classified as matches (red) if the observed m/z corresponds to expected fragments (shifted or unshifted due to metabolism) or as mismatches (blue) if the observed m/z is unshifted but the fragment contains a chemical modification that would alter the m/z. Peaks not used for structure assignment are displayed in black [15]. Metabolite‐based fragments: Fragments are also generated directly from the metabolite structure after predefined metabolic transformations. These fragments, referred to as metmatches (orange), are assigned to the metabolite spectra to capture peaks/structural fragments not derived from the parent molecule. This fragmentation strategy is particularly beneficial for cyclic peptides, where metabolites may come from ring‐opening via amide hydrolysis metabolic reaction, resulting in fragmentation patterns markedly different from the parent [15]. Metabolite‐based fragments: Fragments are also generated directly from the metabolite structure after predefined metabolic transformations. These fragments, referred to as metmatches (orange), are assigned to the metabolite spectra to capture peaks/structural fragments not derived from the parent molecule. This fragmentation strategy is particularly beneficial for cyclic peptides, where metabolites may come from ring‐opening via amide hydrolysis metabolic reaction, resulting in fragmentation patterns markedly different from the parent [15]. Metabolite structures are scored (MMS score) by summing the intensities of fragment assignments for matches and metmatches and subtracting the intensities of mismatches. When multiple metabolite structures are generated with the same m/z, the highest scoring assignments are reported as candidate metabolite structures [16]. This MMS score incorporates fragment assignments from both MS and MS/MS spectra. Additionally, fragments that do not have a unique structural interpretation for a given m/z peak are considered, as are fragments that are present in multiple charge states. After this structure elucidation procedure, manual data interpretation is performed by the user, who evaluates each metabolite candidate using diverse data analysis criteria to systematically filter potential false positive assignments. The criteria used for this annotation are described in detail in the reference study [13] but include mass spectral characteristics (e.g., m/z difference, MS area, negative control area ratios, and fragmentation patterns), as well as chromatographic peak shape and kinetic information, among others. In this context, the MMS score provides a numerical value that supports interpretation, where a high score indicates confidence in the structural elucidation, reflecting a higher number of matching product ions, a lower number of mismatching product ions, and a smaller average m/z deviation between observed and theoretical values. Within this framework, isotopic all similarity is another MetID property used in this study, which quantifies the agreement between observed and theoretical isotopic patterns for each peak. Values range from 0 to 100, where lower values indicate greater deviation from the expected isotopic distribution. It was selected due to its fixed numerical scale, which enables threshold‐based manual data interpretation. This metric is independent of MS/MS fragmentation data and is based solely on the agreement between the observed and theoretical isotopic patterns derived from the molecular formula. As a result, deviations from the expected isotopic distribution can be used as an indicator of potential false positive metabolite assignments. Charge State Handling To account for multiply charged macromolecules, MMS supports the assignment of multiple charge states for each detected molecular species. When the multiple charge ions option is enabled, the software uses the multiple m/z generated with the selected user defined charge range to identify the potential compounds. For peptides, typical charge states range from 3 to 5, whereas for larger macromolecules such as oligonucleotides, charge states up to 15–20 can be considered. The algorithm clusters all ions in the full‐scan chromatogram that share the same retention time peak, thereby grouping different charge states that originate from the same chemical species. At the MS/MS level, fragment coverage accounts for all fragment ions across their allowed charge states. Multiple charge forms of the same fragment are therefore collapsed into a single annotation. Biotransformation Reactions and Fragment Ion Nomenclature for Peptides and Oligonucleotides In the developed approach, two visualization modes are available to represent polymeric compounds such as peptides and oligonucleotides during data analysis [13]. Structures can be displayed either in an expanded form, showing all atoms and intermonomer bonds, or in a nonexpanded (monomer) form, where the polymer is represented by linked monomer acronyms. In this work, the nonexpanded monomer representation is used. Traditionally, fragment ion nomenclature follows the system proposed by Roepstorff and Fohlman, in which ions are classified as a–c (from the N‐terminus) or x–z (from the C‐terminus) [17]. In this scheme, fragment ions are numbered according to their position relative to a well‐defined linear sequence, starting from either terminus. However, this linear‐based nomenclature is not directly applicable to cyclic peptides or peptides with branched architectures, due to the absence of inherent N‐ or C‐termini and the variety of cyclization patterns. As a result, assigning fragment ion numbers according to conventional rules becomes impractical. To overcome this limitation, the nomenclature illustrated in Figure 1A assigns fragment ions based on their position along the peptide backbone, proceeding in a single, defined direction around the cycle. In this system, the number of a fragment ion reflects its sequence position relative to the starting monomer. Consequently, C‐terminal ions are no longer numbered relative to the C‐terminus; for example, the fragment generated between the first and second monomer is denoted x2/y2/z2 rather than x3/y3/z3 as in the conventional nomenclature. This directional numbering provides a consistent and unambiguous way to describe fragments in cyclic or branched peptides, without relying on conventional linear sequence numbering. (A) Proposed fragment ion nomenclature for peptides. (B) Fragment ion nomenclature for oligonucleotides according to McLuckey. For oligonucleotides, the monomer breaking rules include phosphodiester and phosphoramidate bond cleavage. Fragment ion nomenclature follows the system proposed by McLuckey et al. [18]. In this scheme, cleavage of the backbone yields a–d ions containing the 5′‐OH group and w–z for fragments containing the 3′‐OH group (Figure 1B). These cleavage rules and nomenclature are used to interpret MS/MS spectra and assignment of fragment ions to each monomer in the macromolecule for fragment coverage visualization. Fragment Viewer The fragment viewer tool generates an individual coverage map for the parent compound and for each detected metabolite from LC‐MS/MS data in macromolecule MetID studies. In each case, all annotated fragment ions from the corresponding MS/MS spectrum (Figure 2A) are projected onto the molecular structure, generating an integrated representation of the experimental fragmentation evidence, as illustrated in Figure 2B. This facilitates interpretation of MS/MS data by providing a direct structural representation of all assigned fragment ions, reducing the need for manual inspection of individual MS/MS peaks and their corresponding annotations. (A) MS/MS spectrum of metabolite M13‐1396 (RT = 5.27 min) from taspoglutide, with manually annotated fragment ions assigned to each peak; peak‐by‐peak inspection is required for metabolite structure visualization. Peak colors: black, unassigned peaks; red, matching fragment assignments supporting the proposed metabolite structure; cyan, mismatching fragment; coral, metabolite specific matching fragments without a corresponding substrate fragment. (B) Fragment viewer tool, providing a unified visualization of all observed fragment ions mapped onto the parent metabolite structure. In addition, two complementary coverage–based metrics quantify MS/MS fragmentation of the parent and metabolite structures. Fragment coverage is defined as the proportion of inter‐monomer bonds for which fragmentation is supported by at least one annotated fragment ion, relative to the total number of intermonomer bonds present in the structure (Equation 1). In this metric, each covered inter‐monomer bond is counted only once, regardless of how many fragment ions support it. Complete fragment coverage is defined as the proportion of experimentally observed fragment ions relative to the total theoretical fragment ion space of a structure (Equation 2). This theoretical space is constructed from all possible sets of ion types considered for a given experiment (e.g., a–c and x–z ions in peptide). For each inter‐monomer bond, all ion types defined in the acquisition settings are considered (Figure S1), such that the total number of possible fragment ions is given by the number of inter‐monomer bonds multiplied by the number of ion types. In this metric, each fragment ion type is counted independently. For example, in the macromolecule fragmentation rules, if only a/x ion types are enabled, the number of ion types is (Niontypes=2). If all ion types are considered (e.g., a–c and x–z), then (Niontypes=6). In this study, all peptide and oligonucleotide fragmentation rules were enabled to allow for comparison of metabolites across all datasets. The fragment ions used for fragment coverage visualization are derived from MS/MS spectra rather than full‐scan MS. In data‐dependent acquisition (DDA), a single precursor is isolated and fragmented, allowing fragments to be confidently attributed to the corresponding metabolite. For peptides, the only fragment ions not visualized in the fragment viewer are those produced at disulfide bonds and fragments that do not have a unique structure interpretation for a given m/z peak. For example, in a somatostatin‐derived metabolite (Figure S2), the fragment with m/z 120.0808 could correspond to different phenylalanine (F) amino acids present in the molecule, whereas the fragment with m/z 221.1285 could arise from different threonine‐phenylalanine (T‐F) dipeptides. As an example, a calcitonin derived metabolite (M14‐1851, RT = 5.47) is shown in Figure 3. The fragment coverage was 92.9%, calculated as 13 experimentally observed backbone fragments out of 14 possible amide hydrolysis sites (13/14 × 100). The missing coverage corresponds to the absence of fragmentation between Asn3‐Leu4. Fragment viewer showing fragment coverage and complete fragment coverage for a calcitonin derived metabolite. Cysteine (C) residues are highlighted in purple to indicate the presence of a disulfide bond. The complete fragment coverage was 44.0%, calculated as 37 experimentally observed fragment ions divided by the theoretical fragment ion space (14 intermonomer bonds × 6 ion types = 84 possible theoretical fragment ions), yielding 37/84 × 100. Data Two experimental datasets were used in this study: a peptide dataset and an antisense oligonucleotide (ASO) dataset. The peptide dataset comprised 11 peptides, including both linear and cyclic structures and containing natural as well as modified amino acids (Table 1). Seven commercially available peptides (buserelin, calcitonin salmon, goserelin, histrelin, leuprolide, oxytocin, and secretin human) were individually incubated with four proteolytic enzymes: trypsin, chymotrypsin, pancreatic elastase, and pepsin. TABLE 1Summary of the peptide dataset used in this study, including peptide name, molecular weight, structural classification (linear or cyclic), and the corresponding enzymatic incubation conditions used.Compound nameMolecular weight (Da)StructureIncubation conditionBuserelin1238.66LinearTrypsin, chymotrypsin, pancreatic elastase, pepsinCalcitonin salmon3429.71CyclicTrypsin, chymotrypsin, pancreatic elastase, pepsinGoserelin1269.41LinearTrypsin, chymotrypsin, pancreatic elastase, pepsinHistrelin1323.5LinearTrypsin, chymotrypsin, pancreatic elastase, pepsinLeuprolide1209.4LinearTrypsin, chymotrypsin, pancreatic elastase, pepsinOxytocin1007.19CyclicTrypsin, chymotrypsin, pancreatic elastase, pepsinSecretin human3039.41LinearTrypsin, chymotrypsin, pancreatic elastase, pepsinTaspoglutide3338.71LinearDPP‐4Liraglutide3751.20LinearDPP‐4GLP‐13297.68LinearDPP‐4Somatostatin1637.89CyclicHuman serum Summary of the peptide dataset used in this study, including peptide name, molecular weight, structural classification (linear or cyclic), and the corresponding enzymatic incubation conditions used. Taspoglutide, liraglutide, and GLP‐1 were incubated in the presence of dipeptidyl peptidase‐4 (DPP‐4), whereas somatostatin was incubated with human serum. LC‐MS/MS data for buserelin, calcitonin salmon, goserelin, histrelin, leuprolide, oxytocin, and secretin human were acquired using a Thermo Scientific Q‐Exactive Plus mass spectrometer operated in positive electrospray ionization (ESI) mode. Taspoglutide, liraglutide, and GLP‐1 were analyzed using a Thermo Scientific Q Exactive Hybrid Quadrupole‐Orbitrap mass spectrometer operated in positive ESI mode. LC‐MS/MS data for somatostatin were acquired using a Thermo Scientific Q Exactive mass spectrometer operated under the same ionization conditions. All experiments were performed in full‐scan mode with DDA‐MS/MS acquisition. In total, this resulted in 32 independent experiments processed using MMS. Detailed experimental conditions for this dataset are available in the referenced bibliography [19]. The second dataset corresponds to an ASO with the molecular formula C242H307N91O150P94 and a molecular weight of 7633 Da. The ASO contained 25 monomer units and was incubated in human liver tissue. LC‐MS/MS data were acquired using a Thermo Scientific Q Exactive mass spectrometer in negative ESI mode with DDA‐MS/MS acquisition. Results Manual data interpretation of the peptide and oligonucleotide datasets was performed to label detected metabolites as true metabolites or false positives according to established MetID knowledge criteria [13]. In the peptide dataset, 216 metabolites were identified, of which 130 were classified as true metabolites and 86 as false positives. In the oligonucleotide dataset, 39 metabolites were identified, including 29 true metabolites and 10 false positives. The fragment coverage and complete fragment coverage values for all identified metabolites, together with the corresponding peptide, incubation condition, MetID property values, and metabolite classification, have been added to Supporting Information S2. To evaluate how the different MetID properties (including MMS score, fragment coverage, complete fragment coverage, and isotopic all similarity) prioritize true metabolites over false positives, a ranking‐based performance analysis was performed using mean average precision (MAP), as shown in Equation (3): where APi denotes the Average Precision obtained for the ith experiment and N is the total number of experiments [20]. For the 33 experiments, metabolites were ranked independently according to the numerical value of each property, ordering candidates from highest to lowest MMS score, fragment coverage, complete fragment coverage, or isotopic all similarity, respectively. The AP was then calculated within each experiment evaluating how effectively the ranked list places true metabolites above false positives, with higher ranked true metabolites contributing more strongly to the AP value. Finally, the MAP was obtained by averaging AP values across all experiments (Table S1), providing a global measure across both datasets. This approach allows direct comparison between the different MetID properties, where higher MAP values indicate that true metabolites are preferentially ranked near the top of the list, meaning that higher property values are more strongly associated with true metabolite identifications than with false positives. As shown in Table 2, all MetID properties were able to prioritize true metabolites over false positives to a high degree, with MAP values ranging from 0.9090 to 0.9737. Among the properties, complete fragment coverage achieved the highest performance (MAP = 0.9737), followed by fragment coverage (MAP = 0.9676) and MMS score (MAP = 0.9390). Isotopic all similarity showed the lowest performance (MAP = 0.9090), although it still demonstrated strong discriminative ability. TABLE 2MAP values for different MetID properties across all 33 experiments. Higher MAP values indicate a stronger ability to rank true metabolites higher in the candidate lists.MetID propertyMMS scoreFragment coverageComplete fragment coverageIsotopic all similarityMAP0.93900.96760.97370.9090 MAP values for different MetID properties across all 33 experiments. Higher MAP values indicate a stronger ability to rank true metabolites higher in the candidate lists. To assess whether the proposed fragment coverage metrics yielded significantly higher MAP scores than the established MetID properties, one‐sided paired permutation tests were performed using the MAP scores with 10 000 iterations. The p value was computed as the proportion of permutations where the mean paired difference was at least as large as the observed one. The results of the permutation tests are presented in Table 3. TABLE 3Results of one‐sided paired permutation tests (10 000 iterations) comparing the ranking performance of MetID properties across the 33 experiments.Comparison p valueSignificant (p < 0.05)MMS score vs. complete fragment coverage0.0004YesIsotopic all similarity vs. complete fragment coverage0.0088YesIsotopic all similarity vs. fragment coverage0.0223YesMMS score vs. fragment coverage0.0302Yes Results of one‐sided paired permutation tests (10 000 iterations) comparing the ranking performance of MetID properties across the 33 experiments. These results indicate statistically significant differences in ranking performance for all comparisons involving MMS score or isotopic all similarity (p < 0.05). A Holm–Bonferroni correction [21] was additionally computed to account for multiple comparisons, with all comparisons remaining statistically significant after adjustment (Table S2). Figure 4 presents the distribution of the four MetID properties for true metabolites and false positives across the peptide dataset. Although the distribution suggests that complete fragment coverage provides the greatest separation between the two classes, this observation was further evaluated using precision‐recall (PR) curves (Figure 5). Complete fragment coverage achieved the highest average precision (AP = 0.97), followed by fragment coverage (AP = 0.94), MMS score (AP = 0.920), and isotopic all similarity (AP = 0.87). The PR curves support the improved classification performance of complete fragment coverage relative to the other evaluated MetID properties. Distribution of MetID properties (MMS score, isotopic all similarity, fragment coverage, and complete fragment coverage) for true metabolites and false positives across the peptide dataset. Precision–recall curves comparing the ability of MMS score, isotopic all similarity, fragment coverage, and complete fragment coverage for distinguishing true and false metabolite identifications. Discussion The results indicate that high fragment coverage and complete fragment coverage percentages increase confidence in metabolite structural elucidation, as a larger proportion of theoretically possible backbone fragmentations are supported by observed MS/MS fragment ions. In contrast, low fragment coverage values (approaching 0%) are more frequently associated with false positives, which may arise from contamination, background noise, processing errors, or variations in LC‐MS/MS conditions such as chromatography, ionization efficiency, or mass analyzers performance. The MMS score does not follow a fixed numerical scale, as it is inherently relative and experiment dependent. Therefore, high or low values must be interpreted within the context of each experiment rather than as absolute numerical indicators. For example, a metabolite exhibiting both low MMS score and low fragment coverage is more likely to represent a false positive, whereas a moderate MMS score combined with high complete fragment coverage can still indicate a reliable identification. In practice, combining the MMS score with fragment coverage metrics increases confidence in structure elucidation while reducing the risk of misassigning false positives. The fragment coverage distribution graph in Figure 4 shows two false positive metabolites with high fragment coverage values of 83.3% and 100%, both derived from buserelin following trypsin incubation. These examples illustrate that fragment coverage may be inflated in structures with a low number of intermonomer bonds, because only a single fragment ion is required to mark a bond as covered. The first metabolite contains seven monomers and exhibits a fragment coverage of 83.3%, as six of the seven intermonomer bonds are supported by only a single fragment ion. The second metabolite contains three monomers and achieves a fragment coverage of 100%, as all the intermonomer bonds also are supported by just one fragment ion. In such cases, complete fragment coverage provides complementary information by capturing the extent of fragmentation across all ion types associated with each bond, thereby better reflecting the overall richness of the MS/MS evidence, as illustrated in Figures S3 and S4. Beyond metabolite classification, the parent fragment coverage and complete fragment coverage can be used to assess the quality of the MS/MS spectrum. As described in Section 2.1.1, fragment assignment and structure elucidation of metabolites are performed using two complementary strategies, one of which is based on parent derived fragments where peaks are categorized as matches or mismatches, and the MMS score is calculated based on these assignments. This means that when fragment coverage metrics are low, fragmentation is poor, expecting that the corresponding metabolites will exhibit lower MMS scores and a reduced number of assigned fragment peaks in the MS/MS spectrum. Although the fragment coverage metrics were evaluated using both peptide and oligonucleotide datasets, only a single oligonucleotide compound was included. While these results demonstrate the applicability of the proposed metrics to this class of molecules, validation using a broader and more diverse set of oligonucleotides would further strengthen confidence in their generalizability and robustness. Overall, these findings highlight the potential of fragment coverage metrics to support MetID workflows in LC‐MS/MS studies by facilitating the interpretation of MS/MS data and enabling quantitative assessment of fragmentation coverage. Case Studies Antisense Oligonucleotide MetID The first case study corresponds to the ASO compound. Table 4 reports 10 peaks identified as false positives based on the MetID data analysis criteria [13]. Analysis of fragment coverage and complete fragment coverage shows that false positive metabolites are associated with low percentages, empirically observed in this dataset as < 13.6% and complete fragment coverage < 1.6%. TABLE 4MMS score, fragment coverage, complete fragment coverage, and isotopic all similarity values for false positive metabolite candidates identified in the antisense oligonucleotide compound.MetaboliteMMS scoreIsotopic all similarityFragment coverage (%)Complete fragment coverage (%)M8‐4404 (RT = 14.82)258.78500M10‐2486 (RT = 15.87)885.168.900M13‐2348 (RT = 16.10)896.985.75.90.7M16‐2645 (RT = 16.33)288.291.700M20‐1539 (RT = 16.75)977.379.800M23‐1563 (RT = 16.91)321.567.400M34‐753 (RT = 17.57)862.378.913.61.7M35‐446 (RT = 17.79)773.770.613.01.6M36‐424 (RT = 17.89)525.690.313.01.6M38‐464 (RT = 17.92)256.683.900 MMS score, fragment coverage, complete fragment coverage, and isotopic all similarity values for false positive metabolite candidates identified in the antisense oligonucleotide compound. These coverage metrics should not be used in isolation, as metabolite assessment relies on additional evidence such as chromatographic peak shape, peak area, value of isotopic all similarity, difference between observed and calculated m/z, negative control area, and other analytical criteria [13]. However, fragment coverage and complete fragment coverage provide intuitive measures of MS/MS fragmentation that can rapidly highlight candidates requiring further review (Figure 6), without the need for manual inspection of MS/MS spectra. Metabolite M10‐2486 (RT = 15.87) from the antisense oligonucleotide dataset, classified as a false positive. The fragment coverage and complete fragment coverage percentages are shown below the metabolite structure, both indicating values of 0%. Figure 7 illustrates representative MS/MS spectra of metabolites identified as false positives. The spectra exhibit limited fragmentation information, which is reflected in consistently low fragment coverage and complete fragment coverage values. MS/MS spectra of representative metabolites classified as false positives (M8‐4404, M13‐2348, M16‐2645, M23‐1563, M34‐753, and M36‐424), illustrating poor fragmentation patterns characterized by low fragment coverage and complete fragment coverage values. Taspoglutide MetID This second case study corresponds to taspoglutide, a peptide composed of 30 amino acids. After manual analysis of the MetID data, five metabolites were classified as false positives and 10 as true metabolites. False positive metabolites exhibit fragment coverage values ranging from 0% to 25% and complete fragment coverage values from 0% to 4.2%, whereas true metabolites show substantially higher values, with fragment coverage ranging from 45.5% to 100% and complete fragment coverage from 9.1% to 50%. Although the MS/MS spectrum of metabolite M13‐1396 (RT = 5.27 min), which exhibits a MMS score of 2380.3, isotopic all similarity of 97.2, fragment coverage of 100%, and complete fragment coverage of 50%, contains the underlying fragment information (Figure S5), its direct interpretation can be challenging, as it requires peak‐by‐peak inspection to see the assigned fragment ions. To address this limitation, Figure 8 introduces the fragment viewer, which provides an integrated representation of the experimental fragmentation, thereby avoiding the exhaustive manual spectrum inspection. Instead of manually inspecting more than 60 peaks to determine, for example, whether fragment ion y21 is present, this functionality enables rapid verification. Fragment viewer representation of metabolite M13‐1396 from the taspoglutide compound, in which fragment ions from the MS/MS spectrum are mapped onto the metabolite structure. To further illustrate the distinction between fragment coverage and complete fragment coverage, Figure 9 shows the same metabolite (M13‐1396) after restricting the fragmentation rules to the b/y bond. Under these settings, fragment coverage remains 100% because all inter‐monomer bonds are represented by at least one fragment ion. However, complete fragment coverage increases to 75%, indicating that not all the theoretical b‐ and y‐type fragment ions corresponding to those covered bonds are observed experimentally. Fragment viewer representation the taspoglutide‐derived metabolite M13‐1396, with the fragmentation rules restricted to b‐ and y‐type fragment ions in the MMS settings. Conclusions This study introduced two complementary fragment coverage metrics, namely, fragment coverage and complete fragment coverage, as MetID properties/metrics designed to improve the interpretation of MS/MS data and enable its quantitative assessment in peptide and oligonucleotide metabolism studies. Across peptide and oligonucleotide datasets, higher fragment coverage and complete fragment coverage values were associated with true metabolite assignments, where low fragment coverage values (approaching 0%) were more frequently observed in false positive cases. Ranking performance analysis across the four MetID properties showed that complete fragment coverage achieved the highest performance in prioritizing true metabolites, followed by fragment coverage. Both metrics showed statistically significant differences compared with MMS score and isotopic all similarity. Fragment coverage and complete fragment coverage capture complementary aspects of MS/MS fragmentation. Complete fragment coverage provides additional information, as it distinguishes between bond coverage and the completeness of the experimentally observed fragmentation. Beyond metabolite classification, these fragment coverage metrics also provide an indication of MS/MS spectral quality. Overall, the proposed metrics provide complementary quantitative information that can support the interpretation of MS/MS fragmentation patterns and complement existing MetID properties in LC‐MS/MS–based metabolite identification of macromolecules. Author Contributions Paula Cifuentes: conceptualization, investigation, writing – original draft, writing – review and editing, data curation, methodology, validation. Albert Garriga: software. Fabien Fontaine: software, supervision. Ismael Zamora: supervision, project administration. Funding This study was supported by Doctorats Industrials, AGAUR, Generalitat de Catalunya (00002/2023). Conflicts of Interest P.C., F.F., and A.G. are employees of Lead Molecular Design, S.L., and I.Z. is the CEO of the company. Lead Molecular Design, S.L., develops analytical software, including MassMetaSite, which were used in this study. Due to licensing restrictions, the source code of Oniro cannot be distributed directly. However, the software is accessible via a publicly available AWS‐hosted server that includes a free trial option, allowing readers to use the software and reproduce the analyses described in the manuscript (https://oniro.mass‐analytica.com/). Supporting information Figure S1: Macromolecule fragmentation rules defining the set of ion types considered in peptide or oligonucleotide MS/MS experiments. The selection of ion types determines the theoretical fragment ion space used for complete fragment coverage calculations and can be configured in the processing settings. Figure S2: A somatostatin derived metabolite is shown as an example. Peak m/z 120.0808 corresponds to the phenylalanine fragment via amide hydrolysis, but due to the metabolite structure, it could match three fragment pairs: [a6, y6], [a7, y7], and [a11, y11]. Peak m/z 221.1281 corresponds to the dipeptide T‐F fragment and could match two fragment pairs: [a11, y10] and [a12, y11]. These ambiguous cases are excluded from fragment viewer and fragment coverage metrics. Figure S3: Fragment viewer displaying fragment coverage and complete fragment coverage for a buserelin‐derived metabolite, illustrating a case in which fragment coverage is inflated because each inter‐monomer bond is supported by only a single fragment ion despite the limited fragmentation evidence. Figure S4: Fragment viewer displaying fragment coverage and complete fragment coverage for a buserelin‐derived metabolite, highlighting the additional structural information provided by complete fragment coverage in metabolites with a low number of inter‐monomer bonds. Figure S5: MS/MS spectra of taspoglutide and its metabolite M13–1396. The top panel shows the parent peptide, and the bottom panel shows the corresponding metabolite spectrum. Table S1: Average Precision (AP) values for MMS Score, Fragment Coverage, Complete Fragment Coverage, and Isotopic All Similarity across all experiments. Table S2: Results of the Holm–Bonferroni multiple testing correction applied to the one‐sided paired permutation tests comparing the ranking performance (MAP). Acknowledgments This work has been partially supported by Doctorats Industrials, AGAUR, Generalitat de Catalunya (Industrial Doctorate Grant 00002/2023). The authors acknowledge the use of ChatGPT (OpenAI; accessed August 2026) to improve the language, grammar, and readability of portions of the manuscript. All AI‐assisted text was reviewed and revised by the authors to ensure accuracy and clarity of meaning. Data Availability Statement The data that support the findings of this study are available in the Supporting Information. References Associated Data Supplementary Materials Figure S1: Macromolecule fragmentation rules defining the set of ion types considered in peptide or oligonucleotide MS/MS experiments. The selection of ion types determines the theoretical fragment ion space used for complete fragment coverage calculations and can be configured in the processing settings. Figure S2: A somatostatin derived metabolite is shown as an example. Peak m/z 120.0808 corresponds to the phenylalanine fragment via amide hydrolysis, but due to the metabolite structure, it could match three fragment pairs: [a6, y6], [a7, y7], and [a11, y11]. Peak m/z 221.1281 corresponds to the dipeptide T‐F fragment and could match two fragment pairs: [a11, y10] and [a12, y11]. These ambiguous cases are excluded from fragment viewer and fragment coverage metrics. Figure S3: Fragment viewer displaying fragment coverage and complete fragment coverage for a buserelin‐derived metabolite, illustrating a case in which fragment coverage is inflated because each inter‐monomer bond is supported by only a single fragment ion despite the limited fragmentation evidence. Figure S4: Fragment viewer displaying fragment coverage and complete fragment coverage for a buserelin‐derived metabolite, highlighting the additional structural information provided by complete fragment coverage in metabolites with a low number of inter‐monomer bonds. Figure S5: MS/MS spectra of taspoglutide and its metabolite M13–1396. The top panel shows the parent peptide, and the bottom panel shows the corresponding metabolite spectrum. Table S1: Average Precision (AP) values for MMS Score, Fragment Coverage, Complete Fragment Coverage, and Isotopic All Similarity across all experiments. Table S2: Results of the Holm–Bonferroni multiple testing correction applied to the one‐sided paired permutation tests comparing the ranking performance (MAP). Data Availability Statement The data that support the findings of this study are available in the Supporting Information.