- Protein identification and peptide-spectrum matching (PSM) are fundamental processes in mass spectrometry-based proteomics. They enable researchers to determine which proteins are present in a biological sample by identifying the peptide sequences generated from enzymatically digested proteins. In modern proteomics, liquid chromatography–tandem mass spectrometry (LC-MS/MS) is widely used to separate peptides, measure their mass-to-charge ratios, fragment selected ions, and generate tandem mass spectra. Bioinformatics tools then compare these experimental spectra with theoretical spectra predicted from protein sequence databases. Through peptide identification, protein inference, statistical validation, and functional annotation, researchers can characterize the protein composition of cells, tissues, organisms, and biological fluids.
- Protein identification generally begins with the extraction of proteins from a biological sample. Samples may include cells, tissues, blood, microorganisms, plants, animal products, or other biological materials. Proteins are extracted using suitable buffers and processed to remove interfering substances. They are commonly denatured to unfold their structures, reduced to break disulfide bonds, and alkylated to prevent disulfide bonds from reforming. Proteins are then digested into smaller peptides using proteolytic enzymes. Trypsin is the most widely used enzyme because it usually cleaves proteins after lysine and arginine residues, producing peptides suitable for mass spectrometry. Other enzymes, including Lys-C, Glu-C, chymotrypsin, and Asp-N, may be used when alternative cleavage patterns are required.
- The resulting peptide mixture is separated by liquid chromatography, usually reversed-phase chromatography, before entering the mass spectrometer. In electrospray ionization (ESI), peptides acquire electrical charges and enter the instrument as gas-phase ions. The mass spectrometer first records an MS1 spectrum containing precursor-ion signals, which represent the mass-to-charge ratios and intensities of detected peptide ions. Selected precursor ions are subsequently isolated and fragmented to produce MS/MS spectra. Fragmentation techniques such as collision-induced dissociation (CID) and higher-energy collisional dissociation (HCD) commonly generate peptide fragments that provide sequence information. Electron-transfer dissociation (ETD) and electron-capture dissociation (ECD) are also used, particularly for certain peptides and post-translationally modified proteins.
- Peptide-spectrum matching is the computational process of comparing an experimentally measured MS/MS spectrum with theoretical spectra predicted for candidate peptide sequences. The purpose is to identify the peptide sequence that most plausibly explains the observed fragment-ion pattern. Experimental spectra contain information about fragment-ion masses and intensities, while theoretical spectra are generated from peptide sequences using knowledge of the fragmentation behavior of peptides. A successful match should show agreement between the observed and predicted fragment-ion signals, within appropriate mass tolerances and other scoring criteria.
- In collision-based fragmentation, peptide bonds can break at different positions along the amino acid sequence. The resulting fragment ions are commonly described using the b-ion and y-ion nomenclature. A b-ion retains the N-terminal portion of the peptide, whereas a y-ion retains the C-terminal portion. For example, fragmentation of a peptide can generate a series of b-ions representing progressively longer N-terminal fragments and y-ions representing progressively longer C-terminal fragments. The mass differences between consecutive fragment ions can reveal the identities of individual amino acid residues. A sufficiently informative series of fragment ions can therefore support reconstruction of the peptide sequence.
- Theoretical spectra are generated by predicting the masses of possible fragment ions from candidate peptide sequences. These predictions may include b- and y-ions, expected charge states, neutral losses, isotopic patterns, and other fragmentation features. Theoretical and experimental spectra are compared using measures such as mass accuracy, matched fragment-ion counts, signal intensity agreement, and sequence coverage. A peptide candidate that explains more of the observed spectrum, with fewer unexplained or contradictory signals, may receive a higher match score. Different search engines use different mathematical models and scoring systems to rank candidate peptide-spectrum matches.
- Database searching is one of the most widely used approaches to protein identification. In this method, experimental MS/MS spectra are compared with theoretical spectra generated from peptides derived from a reference protein sequence database. Common databases include UniProt, RefSeq, and Ensembl. The selected database should represent the organisms and biological materials included in the experiment. It may also include contaminant protein sequences, such as keratins and common laboratory contaminants, to help identify signals originating from sample handling rather than the biological material of interest.
- A database-search workflow begins by converting protein sequences into theoretical peptides through simulated enzymatic digestion. The search engine considers digestion rules, missed cleavages, peptide length, precursor charge, and possible modifications. It then calculates theoretical fragment-ion masses for candidate peptides and compares them with experimental MS/MS spectra. Candidate matches are scored and ranked, producing peptide-spectrum matches that can subsequently be evaluated statistically. The quality of the reference database is important: an incomplete, outdated, or inappropriate database may prevent correct identification or increase the possibility of incorrect matches.
- Search parameters have a major influence on identification results. Enzyme specificity defines the expected cleavage pattern, while the number of allowed missed cleavages accounts for incomplete digestion. Precursor mass tolerance and fragment-ion mass tolerance determine how closely experimental masses must agree with theoretical values. These tolerances depend on instrument resolution, calibration, and acquisition conditions. Fixed modifications are assumed to occur consistently at specified residues, whereas variable modifications are allowed to occur on selected residues in some candidate peptides. Common examples include carbamidomethylation of cysteine as a fixed modification in many workflows, and oxidation of methionine or acetylation of protein N-termini as variable modifications. Phosphorylation, glycosylation, ubiquitination, and other post-translational modifications may also be included when relevant to the research question.
- Searching for too many variable modifications can greatly increase the number of candidate peptides and computational complexity. It may also increase the probability of incorrect matches. Search parameters should therefore reflect the sample preparation protocol and the biological question. Open modification searching and mass-shift analysis can help identify unexpected modifications, while specialized workflows may be needed for complex modifications such as glycosylation and phosphorylation.
- Several search engines and software platforms are used for peptide-spectrum matching and protein identification. Mascot is an established database-search engine that uses probability-based scoring. SEQUEST compares experimental spectra with theoretical fragment-ion patterns, while Comet is an open-source search engine derived from the SEQUEST approach. X!Tandem and MS-GF+ provide additional methods for scoring peptide-spectrum matches. MaxQuant incorporates the Andromeda search engine and supports large-scale protein identification and quantification. Proteome Discoverer provides a platform for combining different identification and quantification workflows, while FragPipe integrates tools such as MSFragger for fast database searching. OpenMS provides an open-source framework for mass spectrometry data processing and analysis.
- A major challenge in protein identification is distinguishing correct peptide-spectrum matches from incorrect ones. Because large protein databases can generate many candidate sequences, some incorrect peptides may match experimental spectra by chance. Search engines therefore use statistical scoring and validation procedures to estimate identification confidence. A high match score alone does not guarantee that a peptide identification is correct. The reliability of a result must be assessed using appropriate statistical methods and quality-control thresholds.
- False discovery rate (FDR) estimation is a standard method for controlling incorrect peptide and protein identifications. A common strategy is the target-decoy approach. In this method, the search database contains genuine target protein sequences and artificial decoy sequences, often generated by reversing or shuffling target sequences. Experimental spectra are searched against both sets. Since decoy sequences are not expected to represent genuine proteins in the sample, their matches provide an estimate of the number of incorrect identifications. Target and decoy match counts are then used to estimate the false discovery rate at peptide-spectrum match, peptide, and protein levels.
- A simplified expression for false discovery rate is:
- FDR = Estimated false identifications / Total accepted identifications × 100
- This expression illustrates the general concept, although practical target-decoy estimators may include correction factors and depend on the specific search strategy. FDR thresholds such as 1% are commonly used for large-scale peptide and protein identification. Importantly, FDR is a property of a set of accepted identifications rather than a direct probability that an individual identification is incorrect. A 1% FDR threshold does not mean that every reported peptide has exactly a 99% probability of being correct.
- Peptide identification and protein identification are related but distinct stages. Mass spectrometry directly provides evidence for peptide sequences, while the proteins from which those peptides originated must be inferred. This is known as the protein inference problem. A peptide may be unique to one protein, shared by several proteins, or present in multiple protein isoforms. Unique peptides provide stronger evidence for distinguishing individual proteins, whereas shared peptides may support a protein group without identifying one specific member. Protein inference algorithms use peptide evidence, sequence relationships, and statistical models to determine which proteins or protein groups are most likely to be present.
- Protein databases frequently contain related protein sequences, alternative isoforms, homologous proteins, and redundant entries. As a result, different proteins may share many peptide sequences. A protein group may therefore contain several proteins that cannot be distinguished using the available peptide evidence. Researchers should interpret protein groups carefully and avoid claiming that a specific isoform has been identified when the observed peptides are not unique to that isoform. The use of appropriate protein databases, unique peptide evidence, and transparent protein-inference methods improves the reliability of reported results.
- Protein identification can also be influenced by post-translational modifications (PTMs). These modifications alter the chemical properties and masses of peptides and can affect their fragmentation behavior. Phosphorylation, glycosylation, acetylation, methylation, ubiquitination, and oxidation are important examples. When PTMs are not included in the search parameters or handled by suitable algorithms, modified peptides may remain unidentified or may be assigned incorrect sequences. Specialized enrichment methods and modification-aware search strategies are frequently used to improve the detection of particular modified peptides. For example, phosphopeptide enrichment can increase the detection of low-abundance phosphorylated peptides, while glycoproteomics workflows can identify glycosylation sites and associated glycan structures.
- De novo peptide sequencing is an alternative to conventional database searching. Instead of comparing an experimental spectrum with a predefined protein database, de novo algorithms infer the amino acid sequence directly from fragment-ion mass differences. This approach is particularly useful when analyzing organisms with incomplete genome or protein sequence databases, novel proteins, or unexpected sequence variants. However, de novo sequencing can be difficult when spectra are incomplete, noisy, or contain ambiguous mass differences. It is therefore often combined with database searching and other validation approaches. De novo results may also be used to discover sequence variants and improve reference databases.
- The quality of peptide-spectrum matching depends on several experimental and computational factors. High-resolution mass spectrometers provide accurate precursor and fragment-ion measurements, improving the discrimination of candidate peptide sequences. Appropriate chromatography helps separate peptides and reduce spectral complexity. Efficient protein digestion, sample cleanup, and reproducible sample preparation improve the quality of acquired spectra. Computational factors include database completeness, correct search parameters, suitable fragmentation models, accurate mass calibration, modification handling, and statistical validation. Low-quality spectra, co-isolation of multiple precursors, contamination, and incorrect database annotations can all reduce identification confidence.
- Protein identification is also closely connected to quantitative proteomics. Once peptides have been identified, their intensities, spectral counts, or reporter-ion signals can be used to estimate protein abundance. Quantification may use label-free approaches or isotope-labeling strategies such as SILAC, TMT, and iTRAQ. Data-dependent acquisition (DDA) and data-independent acquisition (DIA) provide different ways of collecting MS/MS information. DDA typically selects individual precursor ions for fragmentation, while DIA systematically fragments groups of precursor ions within defined mass windows. These acquisition methods influence how spectra are generated and how peptide identification and quantification are performed computationally.
- After proteins have been identified, functional annotation helps interpret their biological roles. Gene Ontology (GO) provides information about molecular functions, biological processes, and cellular components. KEGG and Reactome support pathway analysis, while STRING, BioGRID, and IntAct provide information about protein associations and interactions. Protein sequence and structure resources, including UniProt and the Protein Data Bank (PDB), provide additional information about protein properties and functions. These resources enable researchers to move from lists of identified proteins to biological interpretation, pathway analysis, and molecular network investigation.
- Protein identification and peptide-spectrum matching have applications across medicine, pharmaceutical research, microbiology, agriculture, and animal science. In biomedical research, they support the identification of disease-associated proteins, biomarkers, and molecular pathways. In pharmaceutical studies, they help investigate drug targets, therapeutic responses, and mechanisms of toxicity. In microbiology, they enable characterization of microbial proteomes, host–pathogen interactions, and changes in protein expression under different environmental conditions. In animal science, they can identify proteins associated with growth, feed efficiency, muscle development, milk production, egg quality, reproduction, immune responses, and adaptation to environmental stress. They can also contribute to the study of genetic variation and its effects on protein sequences and functions.
- Despite major advances in mass spectrometry and bioinformatics, protein identification remains challenging. Low-abundance proteins may be difficult to detect in complex samples, while highly abundant proteins can dominate the measured signals. Shared peptides complicate protein inference, and incomplete databases can limit the discovery of novel or organism-specific proteins. PTMs and sequence variants can further complicate peptide matching. In addition, incorrect search parameters, inadequate FDR control, and poor-quality spectra can lead to unreliable identifications. Careful experimental design, suitable reference databases, validated computational workflows, biological replication, and transparent reporting are therefore essential.
- Protein identification and peptide-spectrum matching form the analytical foundation of mass spectrometry-based proteomics. By matching experimental MS/MS spectra with theoretical peptide sequences, researchers can identify the peptides and infer the proteins present in complex biological samples. Improvements in high-resolution mass spectrometry, spectral prediction, machine learning, database searching, and statistical validation continue to increase the sensitivity, accuracy, and scale of proteomic analysis. These developments are expanding the ability to investigate biological systems, understand disease mechanisms, and discover protein markers relevant to medicine, agriculture, and animal breeding.