# Sequence Tagging in Proteomics: Using Open Modification Search and Tag-Based Approaches to Identify Modified Peptides

## Scope and Reader Context

Researchers analyzing mass spectrometry proteomics data frequently encounter peptides that carry modifications not specified in their search parameters. Standard database search engines compare acquired tandem mass spectra against theoretical spectra generated from a reference proteome, allowing only a predefined set of modifications such as oxidation of methionine or carbamidomethylation of cysteine. When a peptide carries an unexpected modification, the observed fragment ion masses do not match the theoretical spectrum, and the peptide remains unidentified. This article explains two complementary strategies for recovering such peptides: sequence tagging and open modification search. Both approaches expand the search space beyond fixed modification lists and enable the identification of peptides with unknown or unexpected post-translational modifications. The practical outcome for researchers is a workflow for choosing between these methods, implementing them with available tools, and interpreting results with appropriate caution.

## The Problem of Unknown Modifications in Database Search

### Why Standard Database Searches Miss Modified Peptides

Conventional proteomics database search operates on a closed-world assumption. The search engine receives a list of possible modifications and their mass shifts, then attempts to match each acquired spectrum to a peptide sequence from the database with those modifications applied. This approach works well when the modification is known and included in the search parameters. However, biological samples contain a far wider range of modifications than typical search parameter files specify.

Post-translational modifications expand the chemical diversity of proteins far beyond what the genome encodes. Proteins exist as multiple chemical and sequence-specific proteoforms, each potentially serving distinct physiological or pathological roles, and this diversity arises from alternative splicing, translation into amino acid sequences, and various post-translational modifications that lead to an exponential increase in biological complexity [8]. When a researcher searches data without specifying a particular modification, any peptide carrying that modification will fail to match its theoretical spectrum.

The consequences of missed identifications extend beyond simply losing a few peptides. Modified peptides often carry biologically significant information. For example, covalent modification of proteins by drugs or their reactive metabolites has emerged as an important mechanism underlying pharmacological activity, and discovery-driven chemical proteomics approaches have mapped modification landscapes in vivo to identify target proteins and enriched biological pathways [11]. If the search strategy cannot detect these modifications, the researcher misses the central finding of the experiment.

### The Search Space Explosion Problem

The naive solution to finding unknown modifications is to include all possible modifications in the search. This approach fails for a simple reason: the search space becomes computationally intractable. If a search engine considers every amino acid position in every peptide as a potential modification site, and every possible mass shift as a candidate modification, the number of candidate peptides grows by orders of magnitude. The false discovery rate also rises because more candidate matches increase the chance of random spectrum-peptide matches.

Open modification search and sequence tagging solve this problem through different computational strategies. Open modification search allows a large mass offset between the precursor ion mass and the peptide mass, effectively searching for peptides whose mass differs from the database sequence by any value within a specified range. Sequence tagging uses short stretches of confidently assigned fragment ions to derive partial peptide sequences, called tags, which then serve as anchors for database matching.

## Core Principles of Sequence Tagging

### How Sequence Tags Work

Sequence tagging begins with the observation that even when a peptide carries an unknown modification, many of its fragment ions remain unmodified. The fragment ions that do not include the modified residue produce mass differences corresponding to amino acid residues. By examining the mass differences between consecutive fragment ions in a spectrum, a researcher can derive a short sequence of amino acids called a tag.

A typical tag consists of three to five amino acids. The tag provides two pieces of information: the partial amino acid sequence and the mass offset between the tag and the corresponding region of the database peptide. This offset represents the mass of any modification present within the tagged region. The search engine then uses the tag to filter the database, looking for peptides that contain the tag sequence and whose total mass matches the precursor mass within the allowed tolerance.

The key advantage of sequence tagging is that it does not require prior knowledge of the modification. The tag is derived directly from the spectrum, and the mass offset is calculated instead of assumed. This makes sequence tagging particularly useful for discovering novel modifications or modifications that occur at low stoichiometry.

### Tag-Based Search in Practice

Several software tools implement tag-based searching. The general workflow involves three steps. First, the software extracts candidate tags from each spectrum by analyzing fragment ion mass differences. Second, the software searches the protein database for peptides containing those tags. Third, the software validates the matches by comparing the full theoretical spectrum of each candidate peptide against the acquired spectrum.

The quality of the extracted tags determines the success of the search. Tags derived from noisy spectra or from regions of the peptide with poor fragmentation will produce false matches. Most tag-based tools apply quality filters, requiring a minimum number of consecutive fragment ions and a minimum tag length. Longer tags provide more specificity but are harder to derive from real spectra.

### Limitations of Sequence Tagging

Sequence tagging has several limitations that researchers should understand before choosing this approach. First, the method depends on the presence of unmodified fragment ions. If the modification is labile and causes extensive fragmentation of the peptide backbone, the spectrum may not contain enough consecutive fragment ions to derive a tag. Second, tags are typically short, and short tags match many database peptides by chance. This increases the false discovery rate and requires careful statistical validation. Third, the method requires a complete and accurate protein database. Peptides from proteins not present in the database cannot be identified, regardless of the tag quality.

## Open Modification Search Strategies

### The Open Search Concept

Open modification search takes a different approach to the same problem. Instead of deriving tags from fragment ions, the search engine allows a large mass offset between the measured precursor mass and the theoretical peptide mass. In a conventional search, this offset is limited to a small tolerance, typically a few parts per million. In an open search, the offset can range over hundreds of daltons, covering the mass shifts of most common post-translational modifications.

The search engine considers every peptide in the database as a potential match for every spectrum, regardless of the mass difference between them. The mass difference between the precursor ion and the peptide is then interpreted as the mass of an unknown modification. This approach can identify peptides with any modification whose mass falls within the search window, without requiring prior knowledge of the modification.

### Tools for Open Modification Search

Several software tools implement open modification search. The choice of tool depends on the researcher's computational resources, the size of the dataset, and the desired output format. Some tools are integrated into broader proteomics software suites, while others are standalone programs.

OpenSearch is a tool designed specifically for open modification searching. It uses a fragment-ion indexing strategy to speed up the search, allowing it to handle the large search space created by the open mass window. MODa is another tool that implements open modification search with a different algorithmic approach, using a modification-tolerant database search that can identify peptides with multiple modifications.

When selecting a tool, researchers should consider the following factors: the speed of the search, the memory requirements, the format of the output files, and the compatibility with downstream analysis tools. Some tools produce output in standard formats that can be imported into platforms like Bioconductor for further statistical analysis [3]. Others produce proprietary formats that require the vendor's software for visualization and interpretation.

### Comparing Open Search and Sequence Tagging

Open modification search and sequence tagging have complementary strengths and weaknesses. Open search is generally faster and can handle larger datasets, but it provides less information about the location of the modification within the peptide. The mass offset is assigned to the entire peptide, not to a specific residue. Sequence tagging, by contrast, can localize the modification to the tagged region, providing more structural information.

Open search is also more sensitive for modifications that cause a large mass shift, because the search window covers the entire range of possible shifts. Sequence tagging may miss such modifications if the modified residue falls within the tagged region and disrupts the fragment ion series. However, sequence tagging can identify modifications that open search misses when the precursor mass measurement is inaccurate or when the modification is accompanied by a neutral loss that complicates precursor mass assignment.

## Practical Workflow for Identifying Modified Peptides

### Step 1: Assess the Data and Define the Question

Before choosing a search strategy, the researcher should examine the data to understand the scope of the problem. Start by running a conventional database search with standard modifications. Review the identification rate and the distribution of precursor mass errors. If a substantial fraction of high-quality spectra remain unidentified, and if the precursor mass errors show a systematic offset, the data likely contain unexpected modifications.

Define the biological question clearly. Are you looking for a specific known modification, such as phosphorylation or acetylation? Are you searching for any modification within a mass range? Are you trying to discover novel modifications? The answer determines whether you need open search, sequence tagging, or both.

### Step 2: Prepare the Protein Database

Both open search and sequence tagging require a protein database. The quality of the database directly affects the results. Use a complete, well-annotated database for the organism under study. The National Center for Biotechnology Information provides official descriptions of its databases, search systems, sequence resources, and analysis services, and researchers should consult these resources to select the appropriate database for their organism [1].

For human samples, the UniProt human proteome is a common choice. For model organisms, the corresponding reference proteome should be used. If the sample may contain contaminants or proteins from other organisms, add those sequences to the database. A common practice is to append a list of common contaminants, such as keratins and trypsin, to the database.

### Step 3: Configure the Search Parameters

The search parameters for open modification search differ from those for conventional search. The precursor mass tolerance must be set to a large value, typically several hundred daltons. The fragment ion tolerance remains small, usually 10 to 20 parts per million for high-resolution instruments. The enzyme specificity and missed cleavage settings should match the sample preparation protocol.

For sequence tagging, the parameters include the minimum tag length, the maximum number of tags per spectrum, and the mass tolerance for tag matching. These parameters affect the sensitivity and specificity of the search. Longer tags reduce false matches but may miss spectra with poor fragmentation. Start with a minimum tag length of three amino acids and adjust based on the results.

### Step 4: Run the Search and Evaluate the Results

Run the open search or sequence tagging search on the dataset. Examine the distribution of mass offsets in the open search results. A histogram of the mass offsets will show peaks corresponding to common modifications. For example, a peak at +79.97 daltons indicates phosphorylation, and a peak at +42.01 daltons indicates acetylation. These peaks provide a quick overview of the modification landscape in the sample.

For sequence tagging results, examine the identified tags and their corresponding peptides. Check whether the modifications are localized to specific residues and whether the localization is consistent across multiple spectra. Use a localization scoring tool if available to assess the confidence of the modification site assignment.

### Step 5: Validate the Identifications

Validation is a critical step in any modified peptide identification workflow. The false discovery rate for open search and sequence tagging is typically higher than for conventional search, because the larger search space increases the chance of random matches. Apply a target-decoy strategy to estimate the false discovery rate, and set a threshold that balances sensitivity and specificity.

Manual validation of a subset of identifications is recommended. Examine the fragment ion assignments for each identified peptide, paying attention to whether the fragment ions that define the modification site are present and whether they are assigned correctly. For novel modifications, consider synthesizing the modified peptide and analyzing it by mass spectrometry to confirm the identification.

## At a Glance: Choosing Between Open Search and Sequence Tagging

| Consideration | Open Modification Search | Sequence Tagging |
| --- | --- | --- |
| Prior knowledge of modification | None required, any mass shift within window | None required, tag derived from spectrum |
| Modification localization | Not localized, mass offset assigned to whole peptide | Localized to tagged region, may pinpoint modified residue |
| Computational cost | Higher, large search space requires indexing | Lower, tags filter database before full search |
| Sensitivity for large mass shifts | High, window covers full range | Lower, large shifts may disrupt tag extraction |
| Best use case | Discovery of unexpected modifications in large datasets | Localizing known or unknown modifications to specific residues |
| Typical tools | OpenSearch, MODa | Tag-based search tools integrated into proteomics platforms |

## Data Inputs and Quality Considerations

### Required Data for Modified Peptide Identification

The quality of the input data determines the success of any modified peptide identification strategy. High-resolution mass spectrometry data with accurate precursor and fragment ion masses are essential. Instruments with resolution of at least 30,000 for precursor ions and 15,000 for fragment ions provide the mass accuracy needed to distinguish modifications with similar mass shifts.

The fragmentation method also matters. Higher-energy collisional dissociation and electron-transfer dissociation produce complementary fragment ion series. For open search, higher-energy collisional dissociation data are generally sufficient. For sequence tagging, the quality of the fragment ion series is critical, and data with dense, consecutive fragment ions produce better tags.

### Sample Preparation Effects

Sample preparation introduces modifications that can complicate the analysis. Alkylation of cysteine residues with iodoacetamide adds a fixed carbamidomethylation. Oxidation of methionine occurs during sample handling and is typically included as a variable modification. Other artifacts, such as deamidation of asparagine and glutamine, can occur during prolonged storage or digestion.

These sample-derived modifications should be distinguished from biologically relevant modifications. One approach is to run a conventional search with common sample-derived modifications as variable modifications, then compare the results with the open search or sequence tagging results. Modifications identified by both approaches are more likely to be real, while those identified only by the open search may be artifacts.

### Instrument Calibration and Mass Accuracy

Mass accuracy is a critical parameter for both open search and sequence tagging. In open search, the precursor mass tolerance defines the range of detectable modification masses. If the instrument is poorly calibrated, the mass offsets will be inaccurate, and the modification masses will be wrong. In sequence tagging, the fragment ion mass tolerance determines which mass differences are interpreted as amino acid residues. Poor calibration leads to incorrect tags.

Regular instrument calibration with standard peptides is essential. The calibration should be performed before each batch of samples, and the calibration quality should be recorded. If the mass accuracy degrades during the run, the data should be re-calibrated or the affected spectra should be excluded from the analysis.

## Records and Measurements for Reproducible Analysis

### Documenting Search Parameters

Reproducibility requires complete documentation of all search parameters. Record the software version, the database version and source, the enzyme specificity, the number of missed cleavages, the precursor and fragment ion tolerances, the modification list, and any filtering thresholds. This information should be included in the methods section of any publication reporting the results.

The Galaxy Training Network provides accessible workflow training and analysis tutorials that emphasize reproducibility in bioinformatics analysis [4]. Researchers can use these resources to structure their analysis workflows and ensure that all parameters are documented. Similarly, the nf-core documentation describes community pipeline standards for usage, configuration, and reproducible workflow context, which can serve as a model for documenting proteomics analyses [5].

### Recording Quality Metrics

Track quality metrics throughout the analysis. The number of spectra acquired, the number of spectra identified, the identification rate, the false discovery rate, and the distribution of mass offsets should all be recorded. These metrics provide a basis for comparing different search strategies and for troubleshooting when results are unexpected.

For open search, record the histogram of mass offsets and the number of identifications at each offset. This histogram reveals the modification landscape and can be compared across samples to identify condition-specific modifications. For sequence tagging, record the distribution of tag lengths and the number of tags per spectrum. These metrics indicate the quality of the fragmentation data and the sensitivity of the tag extraction.

### Storing Raw Data and Results

Raw data files should be stored in a stable format with complete metadata. The metadata should include the instrument settings, the acquisition date, the sample identifier, and any relevant experimental conditions. Processed results should be stored in a format that can be re-analyzed, such as the standard output formats from the search tools.

The Carpentries Lessons provide foundational training in computing, data, shell, Git, and programming that supports reproducible data management practices [6]. Researchers who apply these practices to their proteomics data will find it easier to revisit analyses, share data with collaborators, and respond to reviewer requests for additional information.

## Common Failure Patterns and Troubleshooting

### Low Identification Rates After Open Search

If the open search identifies fewer peptides than expected, several causes are possible. The protein database may be incomplete or incorrect for the organism under study. The precursor mass tolerance may be too narrow, excluding modifications with large mass shifts. The fragment ion tolerance may be too tight, rejecting valid matches due to small mass errors.

Check the database first. Verify that the database contains the expected proteins and that the sequence coverage is adequate. Then examine the precursor mass error distribution. If the errors are systematically offset, the instrument calibration may be incorrect. If the errors are randomly distributed but large, the tolerance may need to be increased.

### Excessive False Positives in Sequence Tagging

Sequence tagging can produce many false positives when the tags are too short or when the database is large. Short tags of three amino acids match many peptides by chance, especially in large databases. The false discovery rate can be reduced by requiring longer tags, by applying a more stringent false discovery rate threshold, or by using a smaller database.

Another cause of false positives is poor spectrum quality. Spectra with low signal-to-noise ratios produce unreliable tags. Filter the spectra before tag extraction, requiring a minimum number of peaks and a minimum total intensity. This filtering removes low-quality spectra that contribute false tags.

### Inconsistent Modification Localization

When the same modification is identified in multiple spectra but localized to different residues, the localization confidence is low. This situation arises when the fragment ions that distinguish between possible modification sites are missing or of low intensity. The modification may be present, but its exact position cannot be determined.

Use a localization scoring algorithm to assess the confidence of each modification site assignment. Report the localization probability for each site and flag identifications with low localization confidence. For biologically important modifications, consider acquiring additional data with complementary fragmentation methods to improve localization.

## Limitations of Open Modification Search and Sequence Tagging

### Incomplete Databases

Both methods depend on the protein database. Peptides from proteins not present in the database cannot be identified, regardless of the search strategy. This limitation is particularly relevant for samples from organisms with incomplete genome annotations or from mixed communities, such as microbiomes.

For samples containing proteins not encoded in reference genomes, such as therapeutic antibodies and nanobodies, standard database search approaches may fail entirely. Direct protein sequencing approaches that combine streamlined sample preparation, artificial intelligence-driven de novo peptide sequencing, and tailored assembly to reconstruct contiguous protein sequences offer an alternative for such samples [9]. These approaches do not require a reference database and can sequence proteins whose sequences are not encoded in the genome.

### Modification Mass Ambiguity

Open search assigns a mass offset to each identified peptide, but this mass offset may correspond to multiple possible modifications. For example, a mass offset of +42.01 daltons could be acetylation, but it could also be a combination of other modifications with the same total mass. The open search cannot distinguish between these possibilities without additional information.

Sequence tagging provides more localization information but still cannot determine the chemical identity of the modification. The tag defines the modified region, and the mass offset defines the modification mass, but the actual chemical structure of the modification remains unknown. Determining the modification identity requires additional experiments, such as targeted mass spectrometry with specific fragmentation methods or comparison with synthetic standards.

### Stoichiometry and Detection Limits

Modified peptides are often present at low stoichiometry, meaning that only a small fraction of the protein carries the modification. These low-abundance modified peptides may fall below the detection limit of the mass spectrometer, especially in complex samples. The identification of a modified peptide does not indicate its abundance, and the absence of a modified peptide does not indicate the absence of the modification.

Quantitative comparisons between conditions require careful experimental design. Label-based quantification methods, such as isobaric labeling with tandem mass tags, can be combined with open search or sequence tagging to compare modification levels across samples. However, the quantification accuracy depends on the quality of the identifications and the completeness of the modification localization.

## Integration with Quantitative Proteomics Workflows

### Combining Open Search with Quantitative Analysis

Open modification search can be integrated into quantitative proteomics workflows. After identifying modified peptides with open search, the researcher can quantify the relative abundance of each modified peptide across conditions. This approach reveals condition-specific changes in modification levels, which may be more biologically informative than the modification identity alone.

For example, a study of cardiovascular disease might use open search to identify post-translational modifications on cardiac proteins, then quantify the changes in modification levels between healthy and diseased samples. The review of cardiovascular post-translationally modified proteome highlights strategies for analyzing proteoforms using mass spectrometry-based proteomics and bioinformatics, and emphasizes the role of uncommon post-translational modifications in cardiovascular disease [8]. Open search provides a way to discover these uncommon modifications, and quantitative analysis provides a way to assess their biological relevance.

### Using Modified Peptide Identifications in Downstream Analysis

The identifications from open search or sequence tagging can be used in downstream bioinformatics analysis. The modified proteins can be subjected to pathway enrichment analysis to identify biological processes affected by the modifications. The modification sites can be mapped to protein domains and structural features to assess their functional impact.

The SPAID database provides an example of how proteomics-based evidence can be integrated with other data types to support biological discovery. SPAID integrates canonical and non-canonical proteins within a two-level evidence framework, with a validated level containing proteins associated with experimentally confirmed epitopes and a proteomics-based level containing proteins identified by mass spectrometry from human samples, further annotated with differential expression patterns, immunogenicity scores, and functional features [7]. This type of integrated resource demonstrates the value of combining modified peptide identifications with other biological data.

### Reproducibility and Workflow Management

Quantitative proteomics workflows should be managed with attention to reproducibility. The nf-core documentation describes community pipeline standards that emphasize reproducibility, and these standards can be applied to proteomics analysis pipelines [5]. The Galaxy Training Network provides accessible workflow training that supports reproducible analysis [4]. Researchers should use these resources to structure their workflows and ensure that all steps are documented and repeatable.

The Bioconductor project provides official package, workflow, installation, and reproducible genomic-analysis documentation that supports statistical analysis of proteomics data [3]. Researchers can use Bioconductor packages to process the output of open search or sequence tagging tools, apply statistical filters, and generate visualizations of the results.

## Professional Escalation Criteria

### When to Seek Specialized Support

Some situations require specialized expertise beyond what a typical proteomics laboratory can provide. If the open search or sequence tagging results are inconsistent with biological expectations, or if the modification landscape is unexpectedly complex, consider consulting with a bioinformatics specialist or a proteomics core facility.

Specific situations that warrant escalation include: the identification of a novel modification that cannot be explained by known biochemistry, the need for custom software development to handle unusual data formats, the requirement for high-throughput analysis of very large datasets, and the need for orthogonal validation of modification identities using synthetic standards or targeted mass spectrometry.

### When to Question the Data

Researchers should question the data when results are too good to be true. An extremely high identification rate, an unusually large number of modified peptides, or a modification mass that does not correspond to any known biochemical process should all raise concerns. These results may indicate a problem with the database, the search parameters, or the instrument calibration.

The EMBL-EBI Training provides bioinformatics learning pathways, data-resource training, and practical analysis education that can help researchers build the skills needed to critically evaluate their results [2]. Researchers who invest in this training will be better equipped to identify problems in their analysis and to make informed decisions about when to escalate.

## Safety and Regulatory Context

### Data Integrity and Reporting Standards

Proteomics data used in regulatory submissions or clinical studies must meet high standards of data integrity. The raw data, processing steps, and results must be documented in a way that allows independent verification. The false discovery rate and other quality metrics must be reported transparently.

For studies that may inform clinical decisions, the validation requirements are more stringent. Novel modifications identified by open search or sequence tagging should be confirmed by orthogonal methods before being used in any clinical context. The modification identity should be confirmed by comparison with synthetic standards, and the biological relevance should be established through functional studies.

### Ethical Use of Bioinformatics Tools

Bioinformatics tools should be used ethically and responsibly. The software used for open search and sequence tagging should be obtained from legitimate sources, and the terms of use should be respected. The data should be handled in accordance with applicable privacy and confidentiality requirements, particularly for human samples.

The NCBI Data Resources provide official descriptions of databases, search systems, sequence resources, and analysis services, and researchers should use these resources in accordance with their stated terms of use [1]. The EMBL-EBI Training provides guidance on the responsible use of bioinformatics resources [2].

## Decision Framework for Selecting Between Open Search and Sequence Tagging

### Defining the Decision Criteria Before Analysis

Researchers often struggle to choose between open modification search and sequence tagging because the decision depends on multiple interacting factors that are specific to each experiment. A structured decision framework helps avoid the common mistake of defaulting to one approach without considering the data characteristics. The framework below organizes the decision around four criteria: the biological question, the data quality, the computational resources, and the downstream analysis requirements.

The first criterion is the biological question. If the goal is to discover any unexpected modification across a large dataset, open modification search is the appropriate starting point because it surveys the entire mass offset landscape without prior assumptions. If the goal is to determine the precise location of a known or suspected modification on a specific peptide, sequence tagging provides the localization information that open search cannot deliver. These two goals are not mutually exclusive, and many experiments benefit from running both approaches in sequence, with open search providing the discovery layer and sequence tagging providing the localization layer.

The second criterion is data quality, specifically the density and completeness of the fragment ion series. Sequence tagging depends on consecutive fragment ions to derive tags of sufficient length. Spectra from instruments with lower resolution, from peptides with poor fragmentation properties, or from samples with high chemical noise may not yield tags of three or more amino acids. Open modification search is more tolerant of sparse fragment ion series because it relies primarily on the precursor mass offset and the overall spectrum match. Researchers should assess the average fragment ion density in their data before committing to a tag-based strategy.

The third criterion is computational resources. Open modification search expands the search space dramatically, and the computational cost scales with the size of the protein database and the width of the mass offset window. Sequence tagging reduces the computational burden by using tags to filter the database before full spectrum matching. Laboratories with limited computing infrastructure or very large datasets may find sequence tagging more practical, while those with access to high-performance computing can accommodate the larger search space of open modification search.

The fourth criterion is the downstream analysis requirements. If the experiment requires quantitative comparison of modification levels across conditions, the output format and the compatibility with quantification tools matter. If the experiment requires localization probabilities for each modification site, the search tool must provide localization scoring. These requirements should be identified before the search is run, not after the results are obtained.

### Applying the Framework in Practice

The decision framework translates into a practical sequence of steps. First, run a conventional database search with standard modifications to establish a baseline identification rate and to assess the overall data quality. Second, examine the precursor mass error distribution from the conventional search. A systematic offset or a broad distribution of mass errors suggests the presence of unexpected modifications. Third, apply the four decision criteria to select the primary search strategy. Fourth, run the selected search and evaluate the results against the expected outcomes. Fifth, if the primary strategy produces ambiguous results, run the complementary strategy on the same data to resolve the ambiguity.

A practical example illustrates the framework. A researcher studying drug-induced protein modifications in serum samples wants to identify proteins covalently modified by a drug and its metabolites. The biological question is discovery oriented, the data come from a high-resolution instrument with dense fragment ion series, the computational resources are moderate, and the downstream analysis requires pathway enrichment. The framework suggests open modification search as the primary strategy because the question is discovery oriented and the mass offsets of drug modifications are unknown. After open search identifies candidate modified proteins, sequence tagging can localize the modifications to specific cysteine or lysine residues, as demonstrated in chemical proteomics studies of covalent drug modifications [11].

### Recording the Decision Rationale

The decision framework should be documented as part of the analysis record. Record the biological question, the data quality metrics, the computational resources available, and the downstream analysis requirements. Record the rationale for choosing the primary search strategy and the criteria for deciding when to run the complementary strategy. This documentation supports reproducibility and provides a basis for troubleshooting if the results are unexpected.

The Galaxy Training Network provides accessible workflow training that emphasizes reproducibility in bioinformatics analysis, and researchers can use these resources to structure their decision documentation [4]. The nf-core documentation describes community pipeline standards for reproducible workflow context, which can serve as a model for documenting the decision framework [5]. The Carpentries Lessons provide foundational training in data management practices that support this level of documentation [6].

### Common Decision Errors and Their Consequences

Several recurring errors undermine the decision framework. The first error is choosing sequence tagging when the fragment ion series are too sparse to produce reliable tags. The consequence is a low identification rate and a high false discovery rate, because short tags match many database peptides by chance. The remedy is to assess fragment ion density before choosing the strategy and to switch to open modification search if the tag yield is poor.

The second error is choosing open modification search when the biological question requires precise modification localization. The consequence is that the researcher identifies modified peptides but cannot determine which residue carries the modification, requiring a second search with a localization-capable tool. The remedy is to recognize that open search provides discovery, not localization, and to plan for a complementary sequence tagging step when localization is required.

The third error is ignoring the computational cost of open modification search on large datasets. The consequence is excessively long search times or memory exhaustion, which can delay the analysis or require re-running with reduced search parameters. The remedy is to estimate the computational requirements before starting the search and to adjust the database size or the mass offset window accordingly.

The fourth error is failing to document the decision rationale, which undermines reproducibility and makes it difficult to troubleshoot unexpected results. The remedy is to record the decision criteria and the rationale in the analysis log, following the documentation standards described in the nf-core documentation [5].

### Escalation Criteria for the Decision Framework

The decision framework includes specific escalation criteria. If the selected search strategy produces an identification rate below 10 percent of the spectra that passed quality filtering, and the conventional search also produced a low identification rate, the problem may lie in the data quality or the database instead of the search strategy. In this case, consult with a bioinformatics specialist or a proteomics core facility before proceeding.

If the open search identifies a mass offset that appears in a large number of spectra but does not correspond to any known modification, and the offset cannot be explained by a combination of known modifications, the modification may be novel. This situation warrants escalation to a specialized facility with experience in novel modification characterization, including the use of synthetic standards for confirmation.

If the sequence tagging results show inconsistent localization across multiple spectra for the same modification, and the localization probability scores are low, the modification site cannot be confidently assigned. This situation may require additional data acquisition with complementary fragmentation methods, such as electron-transfer dissociation, to improve localization. If the additional data do not resolve the ambiguity, escalate to a specialist.

The EMBL-EBI Training provides bioinformatics learning pathways and practical analysis education that can help researchers build the skills needed to apply this decision framework effectively [2]. Researchers who invest in this training will be better equipped to make informed decisions about search strategy selection and to recognize when escalation is appropriate.

## Frequently Asked Questions

### What is the difference between open modification search and conventional database search?

Conventional database search restricts the precursor mass difference between the measured and theoretical peptide mass to a small tolerance, typically a few parts per million. Open modification search allows a large mass offset, often several hundred daltons, so that peptides with any modification within that mass range can be matched. The mass offset is then interpreted as the mass of an unknown modification.

### How does sequence tagging identify modified peptides without knowing the modification?

Sequence tagging derives short amino acid sequences, called tags, from the mass differences between consecutive fragment ions in a spectrum. These tags are then matched against a protein database. The mass difference between the tag and the corresponding database sequence reveals the mass of any modification within the tagged region, allowing identification without prior knowledge of the modification.

### Which approach should I use for discovering novel modifications?

Open modification search is generally the better first choice for discovering novel modifications because it covers the entire mass range of possible modifications and does not require fragment ion series of sufficient quality for tag extraction. After open search identifies candidate modified peptides, sequence tagging can be used to localize the modification to specific residues.

### How do I estimate the false discovery rate for open search results?

The target-decoy strategy is the standard approach. Create a decoy database by reversing or shuffling the protein sequences, search the data against both the target and decoy databases, and estimate the false discovery rate from the ratio of decoy to target identifications. Set a threshold that balances sensitivity and specificity for your specific research question.

### Can open search identify peptides with multiple modifications?

Yes, open search can identify peptides with multiple modifications, but the interpretation becomes more complex. The mass offset represents the sum of all modification masses on the peptide. Determining which residues carry which modifications requires additional information, such as fragment ions that localize the modifications or complementary data from different fragmentation methods.

### What should I do if my open search identifies a modification mass that does not match any known modification?

First, verify the mass accuracy by checking the instrument calibration and the precursor mass assignment. Then consider whether the mass could represent a combination of known modifications. If the mass remains unexplained, the modification may be novel. Confirm the identification by manual inspection of the spectrum, and consider synthesizing the proposed modified peptide for validation.

### How does database quality affect open search and sequence tagging results?

Database quality is critical for both approaches. Peptides from proteins not present in the database cannot be identified. Incomplete or incorrect databases reduce sensitivity, while redundant or contaminated databases increase the false discovery rate. Use a complete, well-annotated database for the organism under study, and append common contaminants to the database.

### Can I use open search results for quantitative comparisons between samples?

Yes, but with caution. The identification of a modified peptide does not indicate its abundance. For quantitative comparisons, use label-based quantification methods such as isobaric labeling with tandem mass tags, and ensure that the quantification is based on reliable identifications with confident modification localization. The quantification accuracy depends on the quality of the identifications and the completeness of the modification localization.

## Related Bioinformatics Guides

- [Mass Spectrometry-Based Proteomics: Data Analysis Pipelines and Tools](/knowledge/bioinformatics/mass-spectrometry-based-proteomics-data-analysis-pipelines-and-tools)
- [Spatial Proteomics Platforms: A Comparison of Commercial and Open-Source Solutions](/knowledge/bioinformatics/spatial-proteomics-platforms-a-comparison-of-commercial-and-open-source-solutions)
- [Spatial Transcriptomics Methods: A Guide to Experimental Approaches](/knowledge/bioinformatics/spatial-transcriptomics-methods-a-guide-to-experimental-approaches)
- [Bottom-Up Proteomics: Principles, Workflow, and Applications](/knowledge/bioinformatics/bottom-up-proteomics-principles-workflow-and-applications)
- [Pathway Enrichment Analysis for Proteomics: Tools and Interpretation](/knowledge/bioinformatics/pathway-enrichment-analysis-for-proteomics-tools-and-interpretation)

## References and Further Reading

- [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information.
- [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute.
- [Bioconductor](https://bioconductor.org/). Bioconductor Project.
- [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project.
- [nf-core Documentation](https://nf-co.re/docs). nf-core.
- [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries.
- [SPAID: a comprehensive database for disease-specific autoantigens in autoimmune disorders.](https://doi.org/10.1007/s44307-026-00117-8). 2026.
- [Precision Profiling of the Cardiovascular Post-Translationally Modified Proteome.](https://doi.org/10.3390/jcdd13010026). 2026.
- [Generalizable Direct Protein Sequencing With InstaNexus.](https://doi.org/10.1016/j.mcpro.2026.101547). 2026.
- [In Depth Characterization of the Promoter Proximal Proteome of Single Copy Locus FOXP2.](https://doi.org/10.1016/j.mcpro.2026.101570). 2026.
- [Chemical proteomics reveals sinomenine's anti-inflammatory mechanism through serum protein covalent modification.](https://doi.org/10.1186/s13020-026-01403-2). 2026.

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