Mass Spectrometry Protein Identification: From Raw Spectra to Confident Hits

Protein identification by mass spectrometry converts raw peptide fragmentation data into a list of confidently identified proteins through database searching, false discovery rate (FDR) control, and validation. This workflow is central to proteomics research, clinical biomarker discovery, and drug target characterization. For students, researchers, analysts, and life-science professionals, the practical challenge is not simply running a search engine but making defensible decisions about search parameters, database choice, FDR thresholds, and result interpretation. This article provides a practical workflow for protein identification, covering database search, FDR control, and validation, with a decision checklist for choosing search engines and validation parameters.

The Core Workflow: From Raw Spectra to Protein Lists

Protein identification by mass spectrometry follows a defined sequence of steps. The process begins with sample preparation, where proteins are extracted, digested into peptides, and cleaned up before analysis. The resulting peptide mixtures are separated by liquid chromatography and introduced into the mass spectrometer, where precursor ions are selected and fragmented to produce tandem mass spectra. These spectra are then matched against theoretical spectra generated from in silico digestion of a protein database. The final step is protein inference, where peptides identified from tandem mass spectra are assembled into a list of proteins [9].

The quality of the final protein list depends on decisions made at every stage. Sample preparation determines which proteins are present in the measurable pool. Microproteomic workflows, which handle size-limited cell and tissue samples, require specific considerations for sampling, lysis, digestion, and clean-up to reach wide protein dynamic ranges and high numbers of identifications [10]. The choice of acquisition strategy, either data-dependent acquisition (DDA) or data-independent acquisition (DIA), affects the completeness and reproducibility of peptide detection. Database search parameters determine which peptide-spectrum matches are considered. FDR control determines how many false positives are tolerated in the final list.

The workflow is not a single prescriptive pipeline. Different experimental questions require different configurations. A decision-driven framework that emphasizes chemistry-informed hypothesis generation, iterative refinement of candidate modification search space, integration of experimental controls, and targeted data interpretation is more useful than a fixed protocol [13]. This is particularly true when investigating previously uncharacterized protein modifications, where conventional workflows that rely on predefined modification lists may bias discovery toward annotated post-translational modifications [13].

At a Glance: Key Decisions in Protein Identification

The table below summarizes the major decision points in a mass spectrometry-based protein identification workflow, the options available, and the practical considerations for each choice.

Database Selection: The Foundation of Peptide Matching

Peptide identification in most mass spectrometry-based proteomics experiments relies on matching experimental data against peptide and fragment ion masses derived from in silico digests of protein databases [7]. The database is therefore the reference against which all spectra are compared. A database that is too small may miss true proteins. A database that is too large increases search time and the number of false matches.

The standard choice is the canonical proteome of the organism under study, available from resources such as the NCBI Data Resources [2] or EMBL-EBI Training [1]. These databases contain the reference protein sequences for each organism. For most experiments, the canonical proteome is sufficient. However, specific experimental questions may require custom databases.

Proteogenomic approaches integrate proteomic data with genomics and transcriptomics evidence to discover proteins that are absent from canonical databases. In hepatocellular carcinoma research, high-quality Ribo-seq translatomic datasets were used to generate a database of human liver long noncoding RNA-derived open reading frames. Applying this database to proteomics data from tumor-adjacent normal tissue pairs led to the discovery of 104 novel lncRNA-derived microproteins, including 46 differentially expressed between tumor and nontumor tissues and 13 with significant correlation with prognosis [15]. This example demonstrates that the choice of database directly determines which proteins can be identified.

Alternative proteins, produced from unannotated open reading frames, are another class of proteins missed by canonical databases. A combined data-dependent and data-independent acquisition mass spectrometry workflow identified tryptic peptides matching 113 alternative proteins in a panel of 6 pancreatic cancer cell lines [5]. The identification of these proteins required a search space that included unannotated open reading frames. Mining the TCGA and GTEx databases showed that the corresponding transcripts encoding several of these alternative proteins are differentially expressed between pancreatic ductal adenocarcinoma tumors and normal tissues and correlate with patient survival [5].

For clinical and diagnostic applications, the database must also include relevant variants. Antibiotic resistance protein identification in pathogenic bacteria requires a database that contains resistance protein sequences. The MiCId workflow, which performs microorganismal identifications, protein identifications, sample biomass estimates, and antibiotic resistance protein identifications, demonstrated a sensitivity around 85% with a lower bound at about 72% and a precision greater than 95% in identifying antibiotic resistance proteins across Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa samples [12].

Search Engine Selection: Matching Spectra to Peptides

The search engine compares experimental tandem mass spectra against theoretical spectra generated from the database. The choice of search engine affects both the number of identifications and the reliability of those identifications. Different search engines implement different scoring algorithms, different approaches to handling modifications, and different methods for reporting error.

A critical challenge in mass spectrometry proteomics is accurately assessing error control, especially given that software tools employ distinct methods for reporting errors. Many tools are closed-source and poorly documented, leading to inconsistent validation strategies [16]. Three prevalent methods for validating FDR control have been identified: one invalid, one providing only a lower bound, and one valid but under-powered [16]. The result is that the proteomics community has limited insight into actual FDR control effectiveness, especially for DIA analyses [16].

Evaluation of popular DDA tools indicates that these generally seem to control the FDR. However, DIA tools exhibit inconsistent control of the FDR at both the peptide and protein levels, with particularly poor performance on single-cell datasets [16]. This finding has direct practical implications. Researchers using DIA workflows cannot assume that the FDR reported by the software is accurate. They must apply additional validation methods, such as entrapment experiments, to assess actual error control.

Entrapment experiments provide a theoretical framework for rigorously characterizing different FDR control approaches. A more powerful evaluation method has been introduced and applied alongside existing techniques to assess existing tools [16]. This method was first validated in the better-understood DDA setup and then applied to DIA data [16]. For researchers, the practical implication is that entrapment-based validation should be part of the standard workflow when using DIA data, particularly for single-cell datasets where FDR control is least reliable.

Modification Search: Finding What You Did Not Expect

Standard database searches require modifications to be defined in advance. This means that unexpected modifications cannot be identified in a standard setup. Even if high-quality fragment ion spectra of modified peptides were acquired, the search engine will not match them if the modification is not in the search parameters [7].

A stepwise procedure for identifying unexpected modifications uses the database search algorithm Mascot. The workflow includes parallel searches for the identification of known modifications at unexpected amino acids, error tolerant searches for modifications unexpected in the sample but known to the community, and mass tolerant searches for entirely unknown modifications [7]. A follow-up strategy consists of verification of identified modifications in the initial dataset and targeted experiments using synthetic peptides [7].

The decision-driven framework for investigating previously uncharacterized modifications emphasizes chemistry-informed hypothesis generation, iterative refinement of candidate modification search space, integration of experimental controls, and targeted data interpretation [13]. instead of presenting a single prescriptive workflow, this framework highlights key decision points in experimental design, acquisition strategy, and database search configuration that influence confident identification and residue-level localization [13]. The framework is broadly applicable to drug-induced covalent adducts, chemically introduced modifications, and endogenous modifications arising across diverse experimental and biological contexts [13].

Sequence-based modifiers, such as Small Ubiquitin-like Modifier (SUMO)ylation and ubiquitination, present a particular challenge. Most existing search engines are optimized for small, non-fragmenting modifications and struggle to detect large, fragmenting protein-based modifiers [14]. A SUMO-specific search strategy within MaxQuant accounts for the fragmentation behavior of sequence-based modifiers during peptide identification. This approach identified distinct diagnostic features and characteristic mass shifts associated with sequence-based modifier fragmentation, referred to as d-ions (diagnostic ions) and p-ions (PTM ions) [14]. Leveraging these features improved the identification of SUMOylated peptides from human cell lines by approximately 13%, SUMOylation sites in mouse embryonic cells by approximately 22%, and in mouse adipocytes by approximately 24% [14]. The search method improved spectral annotation of sequence-based modifiers by up to 9% increase in the median Andromeda score [14].

False Discovery Rate Control: Separating Signal from Noise

FDR control is the statistical mechanism that limits the proportion of false positives in the final protein list. The concept is straightforward: if a researcher accepts a 1% FDR, then approximately 1% of the identified peptides or proteins are expected to be false positives. The implementation, however, is complicated by the fact that different software tools use different methods for estimating and reporting FDR.

The target-decoy approach is the most common method. A decoy database is generated by reversing or shuffling the protein sequences in the target database. The search engine matches spectra against both target and decoy sequences. The number of matches to decoy sequences provides an estimate of the number of false matches to target sequences. The FDR is then calculated as the ratio of decoy matches to target matches.

The validity of this approach depends on the assumptions made about the distribution of scores for true and false matches. Different tools make different assumptions, and these assumptions affect the accuracy of the reported FDR. The identification of three prevalent methods for validating FDR control, one invalid, one providing only a lower bound, and one valid but under-powered, means that researchers cannot simply trust the FDR reported by their software [16].

For DIA data, the situation is worse. No DIA search tool consistently controls the FDR, with particularly poor performance on single-cell datasets [16]. This finding has significant implications for the growing number of researchers using DIA workflows. The FDR reported by DIA search tools should be treated as an approximation at best. Researchers should apply entrapment-based validation to assess actual error control in their specific dataset.

Entrapment experiments provide a theoretical foundation for evaluating FDR control. In an entrapment experiment, the database is supplemented with sequences that are known to be absent from the sample. Any match to these entrapment sequences is by definition a false positive. The number of entrapment matches provides a direct measure of the false positive rate. This approach allows rigorous characterization of different FDR control methods [16].

Protein Inference: From Peptides to Proteins

Protein inference is one of the most important steps in protein identification. It transforms peptides identified from tandem mass spectra into a list of proteins [9]. The challenge is that a single peptide can match multiple proteins, particularly when those proteins share sequence regions. This occurs with homologous proteins, splice variants, and proteins from the same gene family.

The inference problem is complicated by the fact that some peptides are shared between proteins while others are unique to a single protein. Unique peptides provide strong evidence for the presence of a specific protein. Shared peptides provide evidence for the presence of at least one of the proteins in the group but do not distinguish between them. Protein inference methods must decide how to assign shared peptides to proteins and how to report the resulting protein list.

Different inference methods make different choices. Some methods report only proteins with unique peptides, which is conservative but may miss true proteins. Other methods report all proteins that can be explained by the identified peptides, which is more inclusive but may include false positives. The choice of inference method affects the final protein list and should be made based on the experimental question.

For alternative proteins and microproteins, protein inference is particularly challenging. These proteins are often identified by a small number of peptides, sometimes only one. The evidence for their presence is therefore weaker than for canonical proteins identified by multiple peptides. Proteogenomic approaches that integrate proteomic data with genomics and transcriptomics evidence can strengthen the case for these identifications [15].

Data Acquisition Strategies: DDA and DIA

The choice between data-dependent acquisition and data-independent acquisition affects the nature of the data and the downstream analysis. In DDA, the mass spectrometer selects the most abundant precursor ions for fragmentation. This approach is well-established and produces spectra that are generally well-matched by search engines. DDA tools generally control the FDR [16].

In DIA, the mass spectrometer fragments all precursor ions within a defined mass range, regardless of abundance. This approach provides more comprehensive coverage but produces highly complex spectra that are more difficult to match. DIA tools exhibit inconsistent control of the FDR at both the peptide and protein levels [16]. The complexity of DIA data also makes it more difficult to validate identifications.

The choice between DDA and DIA depends on the experimental question. DDA is appropriate when the goal is to identify the most abundant proteins in a sample. DIA is appropriate when the goal is to achieve comprehensive coverage or to quantify proteins across many samples. The tradeoff is between coverage and reliability. Researchers using DIA must apply additional validation methods to ensure that their identifications are reliable.

A combined DDA and DIA workflow can be used to maximize both coverage and reliability. In the study of alternative proteins in pancreatic cancer cell lines, a combined data-dependent and data-independent acquisition mass spectrometry workflow was used to identify tryptic peptides matching 113 alternative proteins [5]. The combination of acquisition strategies provided complementary information that increased the number of confident identifications.

Specialized Workflows: Imaging, Crosslinking, and Affinity Selection

Mass spectrometry-based protein identification extends beyond standard bottom-up proteomics. Specialized workflows address specific experimental questions and require tailored search strategies.

MALDI imaging mass spectrometry investigates the spatial distributions of thousands of molecules throughout a tissue section from a single experiment. Protein identification is crucial for the biological contextualization of molecular imaging data [6]. However, gas-phase fragmentation efficiency of MALDI generated proteins presents significant challenges, making protein identification directly from tissue difficult [6]. Methods and technologies specifically related to protein identification have been developed to overcome these challenges in MALDI imaging mass spectrometry experiments [6].

Crosslinking mass spectrometry reveals protein-protein interaction sites. An integrated workflow for crosslinking mass spectrometry introduced sequential digestion and the crosslink identification software xiSEARCH. Sequential digestion enhances peptide detection by selective shortening of long tryptic peptides [8]. The workflow includes a simple 12-fraction protocol for crosslinked multi-protein complexes and cell lysates, quantitative analysis, and high-density crosslinking, without requiring specific crosslinker features [8]. This approach reveals dynamic protein-protein interaction sites that are accessible, have fundamental functional relevance, and are suited for the development of small molecule inhibitors [8].

Affinity selection mass spectrometry is used for efficient identification and ranking of potent enzyme inhibitors. The workflow has been optimized for the identification and ranking of potent USP1 inhibitors [17]. This approach combines affinity selection with mass spectrometry detection to identify compounds that bind to a target protein.

Proximity labeling combined with affinity purification-mass spectrometry maps and visualizes protein interaction networks [18]. This workflow enables the identification of proteins in close proximity to a bait protein, providing spatial information about protein interactions.

Membrane protein pharmacology presents specific challenges. Membrane proteins are essential for cellular physiology and the target of half of all FDA-approved drugs. However, their hydrophobicity and low abundance make large-scale expression and purification difficult [11]. Mass spectrometry has enabled workflows for higher-throughput ligand screening, simplified identification of membrane protein targets and ligandable sites, and direct analysis of drug binding in native environments [11]. Emerging mass spectrometry-based strategies include affinity selection, probe-based and probe-free chemoproteomics, and native mass spectrometry [11].

Records and Measurements: Documenting the Workflow

Reproducibility in protein identification requires careful documentation of all workflow parameters. The following records should be maintained for each experiment:

Sample preparation records: protein extraction method, digestion protocol, clean-up procedure, and any deviations from the standard protocol. For microproteomic workflows, the sampling method (flow cytometry, laser capture microdissection) and the specific considerations for size-limited samples should be documented [10].

Acquisition records: mass spectrometer model, acquisition mode (DDA or DIA), chromatographic separation parameters, and instrument settings. These parameters affect the quality and completeness of the data.

Search records: database version and source, search engine and version, fixed and variable modifications, enzyme specificity, precursor and fragment mass tolerances, and FDR thresholds. The database source should be documented, whether from NCBI Data Resources [2], EMBL-EBI Training [1], or a custom database.

Validation records: FDR control method, entrapment database composition, and the results of entrapment experiments. For DIA data, the validation method should be documented explicitly because the software-reported FDR may not be reliable [16].

The FAIR Guiding Principles provide a framework for data management. These principles emphasize that data should be Findable, Accessible, Interoperable, and Reusable [4]. Applying these principles to proteomics data ensures that others can reproduce the analysis and build on the results.

Common Failure Patterns and Troubleshooting

Several failure patterns recur in protein identification workflows. Recognizing these patterns allows researchers to diagnose problems quickly and apply corrective measures.

Low identification numbers: Fewer proteins identified than expected may indicate problems with sample preparation, digestion, or instrument performance. In microproteomic workflows, analyte loss during sample preparation is a common cause of low identification numbers [10]. Specific considerations for processing size-limited samples include optimized lysis, digestion, and clean-up steps [10].

High false discovery rate: More false positives than expected may indicate problems with database selection, search parameters, or FDR control. For DIA data, inconsistent FDR control is a known issue [16]. Entrapment experiments can diagnose the actual FDR and identify whether the software-reported FDR is accurate.

Unexpected modifications missed: Standard searches will not identify unexpected modifications [7]. If the experimental design suggests that modifications may be present, error tolerant and mass tolerant searches should be used [7]. The decision-driven framework for investigating previously uncharacterized modifications provides a structured approach to this problem [13].

Poor reproducibility: Inconsistent results between replicates may indicate problems with sample preparation, acquisition, or data analysis. Standardized protocols and careful documentation of all parameters improve reproducibility.

Protein inference ambiguity: Shared peptides that match multiple proteins create ambiguity in the final protein list [9]. The choice of inference method affects the final list and should be documented.

Limitations and Interpretation Boundaries

Protein identification by mass spectrometry has inherent limitations that affect the interpretation of results. Understanding these limitations is essential for drawing appropriate conclusions.

Identification is probabilistic, not absolute. The output of a search engine is a list of peptide-spectrum matches with associated scores. The FDR provides a statistical estimate of the proportion of false positives, but it does not guarantee that any specific identification is correct. For DIA data, the FDR reported by the software may not accurately reflect the actual error rate [16].

Coverage is incomplete. Mass spectrometry does not identify all proteins in a sample. The dynamic range of protein abundances in biological samples exceeds the dynamic range of mass spectrometry instruments. Low-abundance proteins are often missed. Microproteomic workflows are designed to maximize coverage in size-limited samples but cannot overcome fundamental sensitivity limits [10].

Modifications are under-detected. Standard database searches only identify modifications that are defined in the search parameters [7]. Unexpected modifications are missed even when high-quality spectra of the modified peptides were acquired [7]. Specialized search strategies are required to identify unexpected modifications [7][13][14].

Protein inference is ambiguous. Shared peptides create uncertainty in the assignment of peptides to proteins [9]. The final protein list depends on the inference method used.

Database dependence: The results depend on the completeness and accuracy of the database. Proteins that are absent from the database cannot be identified. Custom databases that include unannotated open reading frames enable the identification of alternative proteins and microproteins [5][15].

Safety and Regulatory Context

Mass spectrometry proteomics involves the use of chemicals, biological samples, and potentially hazardous reagents. Standard laboratory safety practices apply. Sample preparation involves the use of detergents, chaotropes, reducing agents, and enzymes. These reagents should be handled according to their safety data sheets.

For research involving human subjects or clinical samples, data sharing policies apply. The NIH Genomic Data Sharing Policy [3] provides a framework for the sharing of genomic data generated from NIH-funded research. Researchers should be aware of the data sharing requirements that apply to their research.

For clinical applications, the regulatory context depends on the intended use of the results. Biomarker discovery workflows are distinct from clinical diagnostic workflows. Proteomic workflows for biomarker identification require technical and statistical considerations during initial discovery [19]. The transition from discovery to clinical application requires additional validation and regulatory approval.

Professional Escalation Criteria

Researchers should seek expert assistance when they encounter situations that exceed their expertise. The following situations warrant escalation:

Unexpected modifications that cannot be identified with standard search strategies. The stepwise procedure for identifying unexpected modifications requires expertise in search engine configuration and interpretation [7]. If initial searches do not identify the modification, consult a proteomics core facility or collaborator with specialized expertise.

DIA data with inconsistent FDR control. The finding that no DIA search tool consistently controls the FDR [16] means that researchers using DIA data should seek expert advice on validation strategies. Entrapment experiments require careful design and interpretation.

Membrane protein analysis. The hydrophobicity and low abundance of membrane proteins make them challenging for mass spectrometry analysis [11]. Specialized workflows are required, and consultation with experts in membrane protein mass spectrometry is recommended.

Crosslinking mass spectrometry. The identification of crosslinked peptides requires specialized software and expertise [8]. The integrated workflow for crosslinking mass spectrometry provides a starting point, but expert consultation is recommended for complex samples.

Clinical or regulatory applications. The transition from research to clinical application requires expertise in regulatory requirements and validation. Consultation with regulatory experts is recommended.

Frequently Asked Questions

What is the difference between data-dependent acquisition and data-independent acquisition?

Data-dependent acquisition selects the most abundant precursor ions for fragmentation, producing spectra that are generally well-matched by search engines. Data-independent acquisition fragments all precursor ions within a defined mass range, providing more comprehensive coverage but producing more complex spectra. DDA tools generally control the FDR, while DIA tools exhibit inconsistent FDR control at both the peptide and protein levels [16].

How do I choose the right protein database for my experiment?

The standard choice is the canonical proteome of the organism under study, available from NCBI Data Resources [2] or EMBL-EBI Training [1]. Custom databases are needed when the experiment targets alternative proteins, microproteins, or proteins from unannotated open reading frames [5][15]. Proteogenomic approaches that integrate proteomic data with genomics and transcriptomics evidence can accelerate discovery in clinical samples [15].

What is a false discovery rate and why does it matter?

The false discovery rate is the expected proportion of false positives in the identified peptide or protein list. A 1% FDR means that approximately 1% of the identifications are expected to be false. The accuracy of the reported FDR depends on the software tool and the acquisition mode. DIA tools show inconsistent FDR control, so additional validation methods such as entrapment experiments are recommended [16].

How do I identify unexpected protein modifications?

Standard database searches require modifications to be defined in advance, so unexpected modifications are missed [7]. A stepwise procedure includes parallel searches for known modifications at unexpected amino acids, error tolerant searches for modifications unexpected in the sample but known to the community, and mass tolerant searches for entirely unknown modifications [7]. A decision-driven framework emphasizes chemistry-informed hypothesis generation and iterative refinement of the candidate modification search space [13].

What is protein inference and why is it important?

Protein inference transforms peptides identified from tandem mass spectra into a list of proteins [9]. The challenge is that a single peptide can match multiple proteins, particularly when proteins share sequence regions. Different inference methods make different choices about how to assign shared peptides and how to report the resulting protein list [9].

How do I validate my protein identifications?

Validation involves assessing the accuracy of the FDR reported by the search software. Entrapment experiments provide a theoretical framework for rigorously characterizing FDR control [16]. In an entrapment experiment, the database is supplemented with sequences known to be absent from the sample, and any match to these sequences is by definition a false positive. This approach is particularly important for DIA data, where software-reported FDR may not be reliable [16].

What are alternative proteins and how do I identify them?

Alternative proteins are produced from unannotated open reading frames. They have changed our vision of the proteome and have attracted increasing attention from the scientific community [5]. Identifying them requires a database that includes unannotated open reading frames. A combined data-dependent and data-independent acquisition mass spectrometry workflow identified tryptic peptides matching 113 alternative proteins in a panel of 6 pancreatic cancer cell lines [5].

How do I handle size-limited samples in proteomics?

Microproteomic workflows are designed for size-limited cell and tissue samples. Specific considerations include sampling (flow cytometry, laser capture microdissection), sample preparation (lysis, protein extraction, digestion, and clean-up), and analysis (chromatographic or electrophoretic separation, mass spectrometric measurements, and statistical evaluation) [10]. All steps must be optimized to reach wide protein dynamic ranges and high numbers of identifications [10].

Related Bioinformatics Guides

• STRING Database and Protein-Protein Interaction Networks
• De Novo Protein Design Using Computational Structures
• AlphaFold and Beyond: Predicting Viral Protein Structures for Antiviral Target Discovery
• Spike Protein Dynamics and Host Range Prediction Using Molecular Dynamics and Machine Learning
• The Protein Data Bank (PDB): Structural Formats, Coordinates, and Archival Validation Standards

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Mass Spectrometry-Based Workflow for the Identification and Quantification of Alternative and Canonical Proteins in Pancreatic Cancer Cells.. Cells, 2024.
• Protein identification strategies in MALDI imaging mass spectrometry: a brief review.. Current opinion in chemical biology, 2019.
• Identification of Unexpected Protein Modifications by Mass Spectrometry-Based Proteomics.. Methods in molecular biology (Clifton, N.J.), 2019.
• An integrated workflow for crosslinking mass spectrometry.. Molecular systems biology, 2019.
• Protein Inference.. Advances in experimental medicine and biology, 2016.
• Microproteomic sample preparation.. Proteomics, 2021.
• Mass spectrometry-based strategies for membrane protein pharmacology.. Trends in pharmacological sciences, 2025.
• Identification of Antibiotic Resistance Proteins via MiCId's Augmented Workflow. A Mass Spectrometry-Based Proteomics Approach.. Journal of the American Society for Mass Spectrometry, 2022.
• A decision-driven framework for the mass spectrometry analysis of previously uncharacterized protein modifications.. 2026.
• Improved Peptide Search for Identification of SUMO and Sequence-Based Modifiers, in MaxSBM.. 2026.
• A Proteogenomic Approach to Discover Novel lncRNA-Derived Microproteins and Their Potential Clinical Utility in Hepatocellular Carcinoma.. 2026.
• Assessment of false discovery rate control in tandem mass spectrometry analysis using entrapment. Nature Methods, 2025.
• Optimizing the affinity selection mass spectrometry workflow for efficient identification and ranking of potent USP1 inhibitors. Slas Technology, 2024.
• Combined proximity labeling and affinity purification-mass spectrometry workflow for mapping and visualizing protein interaction networks. Nature Protocols, 2020.
• Proteomic workflows for biomarker identification using mass spectrometry - Technical and statistical considerations during initial discovery. Proteomes, 2013.
• Identification of Canonical and Alternative Proteins by Mass Spectrometry. Methods in Molecular Biology, 2026.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.