# Why Standard Protein Databases Miss Novel Peptides: The Case for Sample-Specific Proteogenomic Databases

Standard reference protein databases such as UniProt contain curated, canonical protein sequences that represent common isoforms for well-studied organisms. When you search mass spectrometry data against these databases, you restrict peptide identification to sequences that already exist in the database. Novel peptides arising from single-nucleotide variants, alternative splicing, RNA editing, or sample-specific mutations are absent from standard databases and therefore cannot be identified, regardless of how well your mass spectrometer performed. The solution is a sample-specific proteogenomic database built from RNA-seq or genomic data from the same biological sample. This article explains why standard databases fail, how sample-specific databases are constructed, and provides decision criteria for when custom databases are necessary.

The practical outcome for researchers and laboratory professionals is straightforward: you need a workflow that determines whether your sample contains sequence variation relevant to your biological question, and if so, you need to build a custom database before performing peptide identification. This article provides the framework for making that decision, the steps for building sample-specific databases, and the quality controls needed to avoid false discoveries.

## The Core Problem: Database Search Restricts What You Can Find

Mass spectrometry-based proteomics relies on database search algorithms that compare observed tandem mass spectra against theoretical spectra generated from protein sequences in a reference database. The fundamental limitation is that a peptide must be present in the database as a tryptic peptide of a protein sequence for it to be identified. If the actual protein sequence in your sample differs from the reference sequence at any amino acid position, the resulting peptide may not match any database entry, and the spectrum remains unidentified.

This problem is not hypothetical. Genetic variation between individuals, between tumor and normal tissue, and between laboratory strains and reference genomes produces protein sequences that differ from canonical references. The [Spritz proteogenomic database engine](https://pubmed.ncbi.nlm.nih.gov/32967423) paper describes proteoforms as the workhorses of the cell, noting that subtle differences in amino acid sequences or post-translational modifications can change biological function. When you search mass spectrometry data against a standard database, you miss these variant proteoforms entirely.

The scale of this problem depends on your sample type. For human samples, every individual carries millions of genetic variants relative to the reference genome. Most variants fall in non-coding regions, but a substantial number alter protein-coding sequences. For cancer samples, somatic mutations create tumor-specific protein sequences that are absent from any standard database. For non-model organisms, the situation is worse because the reference database itself may be incomplete or derived from a different strain or subspecies than your experimental sample.

### Why Standard Databases Cannot Capture Sample-Specific Variation

Standard databases are designed for general use. They contain canonical sequences, common isoforms, and well-annotated post-translational modifications. They are not designed to capture the genetic variation present in your specific sample. The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) provide access to comprehensive sequence databases, but these are reference collections, not sample-specific resources. A reference database represents the consensus or representative sequence for each gene, not the actual sequence present in your experimental material.

Consider a concrete example. A tumor sample may carry a somatic mutation that changes a single amino acid in a protein. The wild-type peptide spanning that region will be present in the standard database, but the variant peptide will not. If your mass spectrometry data contains spectra from the variant peptide, database search will fail to identify them. The spectra may be assigned to the wild-type peptide with a mismatch, or they may remain unidentified. Either outcome produces incorrect or incomplete results.

The [proteogenomics data integration chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) explains that genome variability affects both the sequence and abundance level of proteins. Single-nucleotide variants and larger structural variants both contribute to protein sequence differences. The chapter emphasizes that accurate prediction of sample-specific protein sequences requires processing DNA or RNA sequencing data through bioinformatics tools before those sequences can be used as a drop-in replacement in existing peptide identification pipelines.

## How Sample-Specific Proteogenomic Databases Work

A sample-specific proteogenomic database is built from the actual genetic information of your sample instead of from a reference genome. The workflow begins with RNA-seq or genomic DNA sequencing data from the same biological material used for mass spectrometry. The sequencing data is processed to identify variants, splice junctions, and other sequence features that differ from the reference. These sample-specific sequences are then translated into protein sequences and appended to or used in place of the standard database for peptide identification.

The [Spritz tool](https://pubmed.ncbi.nlm.nih.gov/32967423) demonstrates this workflow. Spritz takes raw RNA sequencing data as input and automatically executes approximately 20 tools to construct a proteogenomic database. Sequence variations discovered in the RNA-seq data upon comparison to the Ensembl reference genome are annotated on proteins in the database. Post-translational modification annotations are transferred from UniProt. The resulting database contains the correct amino acid sequences for the sample being analyzed.

The key advantage of this approach is that the database contains the actual protein sequences present in your sample, including variant peptides that would be missed by standard database search. The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) demonstrates that sample-specific databases allow identification of variant peptides, modified variant peptides, and variant proteoforms in bottom-up and top-down proteomic data from human cell lines. The paper also shows identification of phosphorylated variant sites in phosphoproteomic data.

### The Proteogenomic Data Integration Pipeline

Building a sample-specific database requires integrating multiple omics layers. The [proteogenomics methods chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) provides an overview of the different omics layers involved in defining the proteome of a living organism. The chapter discusses how genome variability affects protein sequences and abundance, and it introduces various sequencing technologies from a proteogenomics integration perspective.

The pipeline involves several stages. First, you need sequencing data from your sample. RNA-seq provides information about expressed transcripts, including splice variants and RNA editing events. Genomic DNA sequencing provides information about the genetic variants present in the sample. Long-read sequencing technologies offer advantages for resolving complex genomic regions and full-length transcripts, while short-read sequencing provides higher throughput at lower cost. The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) discusses the respective advantages and shortcomings of these technologies for accurate protein variant prediction.

Second, the sequencing data must be processed to identify variants and other sequence features. This involves aligning reads to a reference genome, calling variants, and assembling transcripts. The [Bioconductor project](https://bioconductor.org/) provides packages for genomic analysis, including variant calling and transcript assembly. The [Galaxy Training Network](https://training.galaxyproject.org/) offers accessible tutorials for these analysis steps, making them available to researchers without extensive bioinformatics expertise.

Third, the sample-specific sequences must be translated into protein sequences and formatted for database search. The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) notes that the goal is to obtain accurately predicted sample-specific protein sequences that can be used as a drop-in replacement in existing approaches for peptide and protein identification using popular database search engines such as MSFragger, SearchGUI, and PeptideShaker.

## At a Glance: Standard Database vs. Sample-Specific Database

The decision between using a standard reference database and building a sample-specific proteogenomic database depends on your sample type, biological question, and available data. The following table summarizes the key considerations.

| Factor | Standard Reference Database | Sample-Specific Proteogenomic Database |
|--------|------------------------------|----------------------------------------|
| Sequence source | Curated canonical sequences from reference genomes | RNA-seq or genomic DNA from your actual sample |
| Variant peptide detection | Not possible, variant sequences absent from database | Possible, variants identified from sequencing data are included |
| Post-translational modification annotation | Curated from literature and databases such as UniProt | Transferred from UniProt and supplemented with sample-specific discovery |
| Data requirements | None beyond mass spectrometry data | Matched RNA-seq or genomic DNA from the same biological sample |
| Analysis time and complexity | Minimal, database is ready to use | Substantial, requires bioinformatics processing of sequencing data |
| False discovery risk | Low for canonical peptides, high for missed variant peptides | Controlled with appropriate filtering, risk of false variants from sequencing errors |
| Best use cases | Well-characterized organisms, samples without expected genetic variation | Cancer samples, non-model organisms, individuals with known genetic variants, any sample where novel peptides are expected |

The choice is not always binary. Many workflows use a hybrid approach where the standard database is supplemented with sample-specific sequences. This approach preserves the ability to identify canonical peptides while adding the capacity to detect variant peptides. The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) describes databases that contain both reference sequences and sample-specific variations, providing a practical model for this hybrid strategy.

## When Standard Databases Are Sufficient

Standard databases remain appropriate for many proteomics experiments. If your biological question does not involve genetic variation, if your organism has a well-annotated reference proteome, and if you have no reason to expect novel peptides, a standard database search will produce reliable results.

Standard databases are sufficient when all of the following conditions hold. Your organism has a high-quality reference proteome with complete annotation. Your sample is from a well-characterized laboratory strain or cell line with minimal genetic divergence from the reference. Your biological question focuses on protein abundance changes instead of sequence variation. You are not studying disease states associated with somatic mutations. Your samples do not come from individuals with known genetic variants affecting protein sequences.

For example, a study comparing protein expression between two treatment conditions in the same cell line can use a standard database. The cell line has a defined genome, and the treatment is unlikely to introduce new protein sequences. The biological question is about abundance changes, not sequence variation. In this case, building a sample-specific database would add complexity without improving results.

The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide guidance on when standard databases are appropriate and how to use them effectively. These resources also cover the limitations of standard databases and the situations where additional analysis is needed.

## When Sample-Specific Databases Are Essential

Sample-specific databases become essential when your sample contains protein sequences that differ from the reference. The following scenarios require custom database construction.

Cancer samples are the most common case. Tumors accumulate somatic mutations, including single-nucleotide variants that alter protein sequences. The [longitudinal tumor-specific peptide study](https://doi.org/10.3390/cells15050474) demonstrates this principle in pediatric brain tumors. The study used an integrated proteogenomic workflow combining discovery and targeted mass spectrometry to identify tumor-specific peptides derived from individual tumors. These peptides were absent from standard databases because they arose from tumor-specific mutations. The study identified tumor-specific peptides from resected tumor tissue and matched them against cerebrospinal fluid proteomic datasets to nominate candidate biomarkers for disease monitoring.

Non-model organisms present another case. If your organism lacks a complete reference proteome, or if the reference comes from a different strain or subspecies than your sample, standard databases will miss organism-specific sequences. Building a database from RNA-seq data from your actual sample ensures that the database contains the correct sequences.

Samples from individuals with known genetic variants require custom databases. If your study involves human subjects with known protein-altering variants, or if you are studying a genetic disease caused by a specific mutation, the variant peptide will only be identified if it is included in the database.

Studies of alternative splicing benefit from sample-specific databases. RNA-seq data reveals which splice isoforms are actually expressed in your sample. Standard databases may contain annotated isoforms, but they may not contain the specific combination of exons present in your sample. Building a database from RNA-seq data captures the actual transcript sequences.

### The Pediatric Brain Tumor Example

The [tumor-specific peptide study](https://doi.org/10.3390/cells15050474) provides a concrete example of why sample-specific databases are essential for cancer research. The study analyzed tumor tissue and serial cerebrospinal fluid samples from six pediatric brain tumor patients, including five medulloblastomas and one atypical teratoid/rhabdoid tumor. The researchers used an integrated proteogenomic workflow combining discovery and targeted mass spectrometry.

Tumor-specific peptides were identified from resected tumor tissue and matched against cerebrospinal fluid proteomic datasets to nominate candidate biomarkers. High-confidence peptides were synthesized as isotopically labeled standards and quantified longitudinally using targeted multiple reaction monitoring. Two tumor-specific peptide biomarkers demonstrated robust detection in serial cerebrospinal fluid samples and exhibited temporal concordance with radiographic disease course, declining with treatment response and increasing during disease progression.

This study would have been impossible with standard databases alone. The tumor-specific peptides arose from somatic mutations unique to each patient's tumor. These sequences were absent from any reference database. The researchers had to build sample-specific databases from tumor sequencing data to identify the peptides, then use targeted mass spectrometry to quantify them in cerebrospinal fluid.

## Building a Sample-Specific Database: Practical Workflow

Constructing a sample-specific proteogenomic database requires a defined workflow with quality controls at each stage. The following steps provide a practical framework.

### Step 1: Assess Whether You Need a Custom Database

Before investing time in database construction, assess whether your sample requires it. Review your biological question, sample type, and existing knowledge about genetic variation in your system. If you are working with cancer samples, non-model organisms, or samples from individuals with known variants, proceed with custom database construction. If you are working with well-characterized laboratory strains and your question is about abundance changes, a standard database may suffice.

Document your assessment and the rationale for your decision. This documentation becomes part of your analysis records and supports the validity of your results.

### Step 2: Obtain Matched Sequencing Data

A sample-specific database requires sequencing data from the same biological sample used for mass spectrometry. RNA-seq is the most common choice because it provides information about expressed transcripts, including splice variants and expressed mutations. Genomic DNA sequencing provides information about all genetic variants, including those not expressed at the RNA level.

The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) discusses the advantages and shortcomings of different sequencing technologies. Short-read sequencing provides high throughput and accuracy for variant detection in exonic regions. Long-read sequencing provides full-length transcript information and resolves complex genomic regions, but at higher cost and lower throughput. Your choice depends on your biological question and available resources.

Ensure that the sequencing data comes from the same biological sample or a closely matched sample from the same individual or culture. Mismatched samples introduce errors because the genetic variants in the sequencing sample may differ from those in the mass spectrometry sample.

### Step 3: Process Sequencing Data to Identify Variants

The sequencing data must be processed to identify variants and assemble transcripts. This involves quality control of raw reads, alignment to a reference genome, variant calling, and transcript assembly. The [Galaxy Training Network](https://training.galaxyproject.org/) provides accessible tutorials for these steps, making them available to researchers without extensive bioinformatics experience.

The [Bioconductor project](https://bioconductor.org/) offers packages for genomic analysis, including variant calling, transcript assembly, and annotation. These packages are designed for reproducible analysis and integrate with the broader R ecosystem for statistical analysis.

Quality control is critical at this stage. Sequencing errors can produce false variants that lead to incorrect protein sequences in your database. Apply appropriate quality filters, including base quality scores, mapping quality, and variant allele frequency thresholds. Document your filtering criteria and the number of variants retained at each stage.

### Step 4: Generate Sample-Specific Protein Sequences

The identified variants and assembled transcripts must be translated into protein sequences. This step requires careful handling of the genetic code, including consideration of splice sites, frameshift mutations, and stop codons.

The [Spritz tool](https://pubmed.ncbi.nlm.nih.gov/32967423) automates this process. Spritz automatically sets up and executes approximately 20 tools to construct a proteogenomic database from raw RNA sequencing data. Sequence variations discovered in the RNA-seq data are annotated on proteins in the database, and post-translational modification annotations are transferred from UniProt.

If you are building the database manually, ensure that your translation pipeline correctly handles all genetic features. Test the pipeline on known sequences to verify that it produces correct translations.

### Step 5: Combine Sample-Specific Sequences with Reference Sequences

The sample-specific protein sequences should be combined with reference sequences to create a comprehensive database. This hybrid approach ensures that canonical peptides are still identifiable while adding the capacity to detect variant peptides.

The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) describes databases that contain both reference sequences and sample-specific variations. This approach preserves the ability to identify canonical peptides while adding the capacity to detect variant peptides.

When combining sequences, ensure that protein identifiers are unique and that the database format is compatible with your search engine. Document the composition of the final database, including the number of reference sequences and the number of sample-specific sequences.

### Step 6: Validate the Database

Before searching your mass spectrometry data, validate the database to ensure it contains the expected sequences. Check that known variant peptides are present. Verify that the database does not contain obvious errors such as truncated sequences or incorrect translations.

The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) emphasizes that the goal is to obtain accurately predicted sample-specific protein sequences that can be used as a drop-in replacement in existing approaches for peptide and protein identification. Validation ensures that the database meets this standard.

### Step 7: Search Mass Spectrometry Data and Apply Quality Filters

Search your mass spectrometry data against the sample-specific database using your preferred search engine. The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) mentions MSFragger, SearchGUI, and PeptideShaker as popular database search engines that accept custom databases.

Apply appropriate quality filters to the search results. The false discovery rate should be controlled using target-decoy approaches. Variant peptides should be manually inspected to confirm that the spectra support the identified sequences.

### Step 8: Document and Report

Document the entire database construction process, including the sequencing data used, the variant calling parameters, the filtering criteria, and the final database composition. This documentation is essential for reproducibility and for reviewers to assess the validity of your results.

Report the number of variant peptides identified and the proportion of total identifications that are sample-specific. This information provides context for interpreting your results and helps readers understand the value of the custom database approach.

## Tools and Resources for Proteogenomic Database Construction

Several tools and resources support the construction of sample-specific proteogenomic databases. The choice of tools depends on your bioinformatics expertise, available computing resources, and specific analysis requirements.

The [Spritz tool](https://pubmed.ncbi.nlm.nih.gov/32967423) provides an automated pipeline for generating protein databases annotated with sequence variations and post-translational modifications. Spritz offers a simple graphical user interface for Windows and scripts that can be run on any operating system. The tool automatically sets up and executes approximately 20 tools for constructing a proteogenomic database from raw RNA sequencing data.

The [Bioconductor project](https://bioconductor.org/) provides packages for genomic analysis, including variant calling, transcript assembly, and annotation. These packages are designed for reproducible analysis and integrate with the broader R ecosystem for statistical analysis. Bioconductor packages are particularly useful for researchers who are comfortable with R programming.

The [Galaxy Training Network](https://training.galaxyproject.org/) offers accessible workflow training and analysis tutorials. Galaxy provides a web-based platform for bioinformatics analysis that does not require command-line expertise. The training materials cover variant calling, transcript assembly, and other steps needed for proteogenomic database construction.

The [nf-core documentation](https://nf-co.re/docs) describes community pipeline standards for reproducible bioinformatics workflows. nf-core pipelines are built on the Nextflow workflow manager and provide standardized, tested pipelines for common analyses. These pipelines can be adapted for proteogenomic database construction.

The [Carpentries lessons](https://carpentries.org/lessons) provide foundational training in computing, data analysis, shell, Git, and programming. These skills are essential for researchers who want to build custom bioinformatics pipelines for proteogenomic analysis.

The [EMBL-EBI Training](https://www.ebi.ac.uk/training) resources provide bioinformatics learning pathways and data-resource training. These resources cover the use of major biological databases and analysis tools, providing a foundation for understanding the data resources used in proteogenomic analysis.

The [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/) provide access to comprehensive sequence databases, search systems, and analysis services. NCBI resources are essential for accessing reference genomes, annotated protein sequences, and other data needed for proteogenomic analysis.

## Quality Controls and Reproducibility Considerations

Quality control is essential throughout the proteogenomic database construction process. Errors at any stage can produce incorrect protein sequences that lead to false peptide identifications.

### Sequencing Data Quality

The quality of the sequencing data directly affects the accuracy of the resulting database. Poor quality sequencing data produces false variants that lead to incorrect protein sequences. Apply appropriate quality filters to raw reads, including base quality scores and adapter trimming. Assess the overall quality of the sequencing run using metrics such as read depth, mapping rate, and error rate.

The [Galaxy Training Network](https://training.galaxyproject.org/) provides tutorials on sequencing data quality control, including the use of tools such as FastQC and Trimmomatic. These tools assess read quality and remove low-quality bases and adapter sequences.

### Variant Calling Accuracy

Variant calling is a critical step that requires careful parameter selection and quality filtering. False positive variants produce incorrect protein sequences in the database. False negative variants cause the database to miss actual variant peptides.

Apply appropriate filters to variant calls, including minimum allele frequency, minimum read depth, and mapping quality. Consider using multiple variant callers and retaining only variants identified by multiple tools. The [Bioconductor project](https://bioconductor.org/) provides packages for variant calling and filtering.

### Database Validation

Validate the final database to ensure it contains the expected sequences. Check that known variant peptides are present. Verify that the database does not contain obvious errors such as truncated sequences or incorrect translations.

The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) emphasizes the importance of accurate protein sequence prediction for proteogenomics data integration. Validation ensures that the database meets this standard.

### False Discovery Rate Control

Control the false discovery rate in peptide identification using target-decoy approaches. This involves searching the data against a decoy database containing reversed or shuffled sequences and using the number of decoy hits to estimate the false discovery rate.

Variant peptides should be manually inspected to confirm that the spectra support the identified sequences. This is particularly important for variant peptides because they are often present at lower abundance and may have lower quality spectra.

### Reproducibility

Document all parameters and steps in the database construction process to ensure reproducibility. Use version control for scripts and workflows. The [Carpentries lessons](https://carpentries.org/lessons) provide training in Git and reproducible research practices.

The [nf-core documentation](https://nf-co.re/docs) describes community standards for reproducible bioinformatics workflows. These standards include versioned pipelines, containerized environments, and comprehensive documentation.

## Common Failure Patterns in Proteogenomic Database Construction

Understanding common failure patterns helps researchers avoid mistakes and troubleshoot problems when they occur.

### Failure Pattern 1: Mismatched Sequencing and Mass Spectrometry Samples

The most common failure is using sequencing data from a different sample than the mass spectrometry data. This produces a database that does not contain the actual protein sequences present in the mass spectrometry sample. Variant peptides are missed because the variants in the sequencing sample differ from those in the mass spectrometry sample.

Prevention: Always use sequencing data from the same biological sample or a closely matched sample from the same individual or culture. Document the sample provenance for both sequencing and mass spectrometry data.

### Failure Pattern 2: Inadequate Sequencing Depth

Insufficient sequencing depth leads to false negative variant calls. Variants present in the sample are not detected because they are not covered by enough reads. The resulting database misses variant peptides.

Prevention: Ensure adequate sequencing depth for your sample type and biological question. For RNA-seq, consider the expression level of the genes of interest. Lowly expressed genes require greater sequencing depth to detect variants.

### Failure Pattern 3: Overly Stringent or Lenient Variant Filtering

Overly stringent filtering removes true variants, causing the database to miss variant peptides. Overly lenient filtering retains false variants, causing the database to contain incorrect protein sequences that produce false peptide identifications.

Prevention: Balance sensitivity and specificity in variant filtering. Use appropriate thresholds based on your sequencing depth and error rate. Validate the database to check for expected variants.

### Failure Pattern 4: Incorrect Translation of Variants

Errors in translating variant sequences to protein sequences produce incorrect database entries. Frameshift mutations, stop codons, and splice site variants require careful handling to produce correct protein sequences.

Prevention: Use validated translation tools and test the pipeline on known sequences. Manually inspect translations of variants in genes of interest.

### Failure Pattern 5: Database Format Incompatibility

The final database must be in a format compatible with your search engine. Format errors cause search failures or incorrect results.

Prevention: Test the database with a small subset of mass spectrometry data before running the full analysis. Verify that the search engine accepts the database and produces expected results.

### Failure Pattern 6: Insufficient Quality Control of Mass Spectrometry Data

Poor quality mass spectrometry data produces unreliable identifications regardless of the database used. This is not specific to proteogenomic databases but is a common source of problems.

Prevention: Apply standard quality control measures to mass spectrometry data, including calibration, lock mass correction, and removal of low-quality spectra.

## Limitations of Sample-Specific Proteogenomic Databases

Sample-specific proteogenomic databases have limitations that researchers must understand to interpret results correctly.

### Database Size and Search Speed

Sample-specific databases are larger than standard databases because they contain both reference sequences and sample-specific variations. Larger databases increase search time and may increase the false discovery rate because there are more sequences to match against.

The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) describes databases that contain both reference sequences and sample-specific variations. The increased database size is a tradeoff for the ability to identify variant peptides.

### Incomplete Variant Detection

The database only contains variants that are detected in the sequencing data. Variants that are not detected due to inadequate sequencing depth, low expression, or technical limitations will be absent from the database. The database is sample-specific but not complete.

### RNA-Seq Does Not Capture All Protein Sequences

RNA-seq provides information about expressed transcripts, but not all transcripts are translated into proteins. Some RNA variants may not produce protein products. Conversely, some protein sequences may arise from post-transcriptional or post-translational processes that are not captured by RNA-seq.

### Post-Translational Modifications

Sample-specific databases built from RNA-seq data do not contain information about post-translational modifications. The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) addresses this by transferring post-translational modification annotations from UniProt and allowing modifications to be discovered and added using bottom-up mass spectrometry data and global post-translational modification discovery.

### Interpretation Complexity

Results from sample-specific database searches are more complex to interpret than results from standard database searches. Variant peptides must be validated to confirm that they arise from actual genetic variation instead of sequencing or database errors.

## Multi-Omics Integration and the Broader Context

Proteogenomic databases are part of a broader trend toward multi-omics integration in biological research. The integration of genomic, transcriptomic, proteomic, and other omics data provides a more complete picture of biological systems than any single omics layer alone.

The [multi-omics survival mediation analysis framework](https://doi.org/10.1371/journal.pcbi.1014217) demonstrates the value of integrating multiple omics layers. The SMAHP method handles high-dimensional exposures and mediators, integrates multi-omics data, and provides a statistical framework for identifying causal pathways on survival outcomes. The method was applied to head-and-neck carcinoma proteogenomic data, detecting a gene mediated by a protein that influences survival time.

The [RNA and protein expression discordance study](https://doi.org/10.1007/s12672-026-04470-x) in lung adenocarcinoma illustrates the importance of integrating transcriptomic and proteomic data. The study found that most genes in tumors showed a positive correlation between RNA and protein expression, but some genes showed drastic discordance during tumorigenesis. DNA topoisomerase II alpha showed a drastic correlation change during tumorigenesis, and its gene and protein expression were associated with overall survival in patients with lung adenocarcinoma.

These examples demonstrate that multi-omics integration provides biological insights that cannot be obtained from any single omics layer. Sample-specific proteogenomic databases are a key tool for this integration because they enable the identification of protein sequences that are specific to the sample being studied.

## Professional Escalation Criteria

Researchers should escalate to expert consultation when they encounter situations that exceed their expertise or when standard approaches fail to produce reliable results.

### When to Consult a Bioinformatics Specialist

Consult a bioinformatics specialist when you lack the expertise to process sequencing data and build custom databases. The [Carpentries lessons](https://carpentries.org/lessons) provide foundational training, but complex analyses may require specialized expertise.

Consult a specialist when your variant calling results are ambiguous or when you are unsure about the appropriate filtering criteria. Incorrect variant calling produces incorrect databases and unreliable results.

### When to Consult a Statistical Geneticist

Consult a statistical geneticist when you need to interpret the biological significance of identified variants. Not all variants affect protein function, and distinguishing functional from neutral variants requires specialized knowledge.

### When to Consult a Clinician or Domain Expert

Consult a clinician or domain expert when your results have clinical implications. The [tumor-specific peptide study](https://doi.org/10.3390/cells15050474) demonstrates the potential clinical value of proteogenomic analysis, but clinical interpretation requires specialized expertise.

### When to Repeat the Analysis

Repeat the analysis when you suspect errors in the database construction process. If the database validation reveals problems, if the search results contain unexpected patterns, or if the results are not reproducible, repeat the analysis with corrected parameters.

## Records and Measurements for Proteogenomic Analysis

Maintaining detailed records is essential for reproducible proteogenomic analysis. The following records should be maintained for each analysis.

### Sample Provenance Records

Document the source of each sample, including the organism, tissue type, individual identifier, and collection date. Document the relationship between sequencing and mass spectrometry samples. This information is essential for interpreting results and for troubleshooting problems.

### Sequencing Data Records

Document the sequencing platform, library preparation method, sequencing depth, and quality metrics. Record the version of the reference genome used for alignment and the parameters used for variant calling.

### Database Construction Records

Document the tools and parameters used for each step of database construction. Record the number of variants identified, the filtering criteria applied, and the final database composition. This documentation is essential for reproducibility.

### Mass Spectrometry Data Records

Document the mass spectrometry platform, acquisition method, and data processing parameters. Record the search engine used, the database version, and the false discovery rate thresholds applied.

### Analysis Results Records

Document the number of peptides identified, the number of variant peptides, and the proportion of identifications that are sample-specific. Record the validation steps performed and the results of manual inspection of variant peptides.

## Safety and Ethical Considerations

Proteogenomic analysis involving human samples raises ethical considerations that researchers must address.

### Informed Consent

Ensure that human samples are collected with appropriate informed consent. The consent should cover the intended use of the samples, including genomic and proteomic analysis. The [tumor-specific peptide study](https://doi.org/10.3390/cells15050474) involved pediatric brain tumor patients, highlighting the importance of appropriate consent for vulnerable populations.

### Data Privacy

Genomic data is sensitive personal information. Ensure that genomic and proteomic data are stored securely and that access is restricted to authorized personnel. De-identify samples and data to protect patient privacy.

### Return of Results

Consider whether and how to return individual research results to participants. This is particularly important when the analysis reveals potentially clinically actionable information.

### Data Sharing

Consider data sharing policies and requirements. Many funding agencies and journals require data sharing. Ensure that data sharing complies with consent agreements and privacy regulations.

## Frequently Asked Questions

### What is a sample-specific proteogenomic database?

A sample-specific proteogenomic database is a protein sequence database built from the actual genetic information of your sample instead of from a reference genome. It is constructed by processing RNA-seq or genomic DNA sequencing data from the same biological material used for mass spectrometry, identifying variants and splice junctions that differ from the reference, and translating these sample-specific sequences into protein sequences. The resulting database contains the actual protein sequences present in your sample, including variant peptides that would be missed by standard database search.

### Why do standard protein databases miss novel peptides?

Standard protein databases contain curated canonical sequences that represent common isoforms for well-studied organisms. They do not contain sequences arising from single-nucleotide variants, alternative splicing, RNA editing, or sample-specific mutations. When your sample contains protein sequences that differ from the reference, the resulting peptides are absent from the database and cannot be identified by database search. The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) explains that subtle differences in amino acid sequences can change biological function, and these differences are missed when searching against standard databases.

### How do I know if I need a sample-specific database?

You need a sample-specific database when your sample contains protein sequences that differ from the reference. This includes cancer samples with somatic mutations, non-model organisms without complete reference proteomes, samples from individuals with known genetic variants, and studies of alternative splicing. If your organism has a well-annotated reference proteome, your sample is from a well-characterized laboratory strain, and your biological question focuses on abundance changes instead of sequence variation, a standard database may be sufficient.

### What sequencing data do I need for building a sample-specific database?

RNA-seq is the most common choice because it provides information about expressed transcripts, including splice variants and expressed mutations. Genomic DNA sequencing provides information about all genetic variants, including those not expressed at the RNA level. The [proteogenomics chapter](https://pubmed.ncbi.nlm.nih.gov/34905178) discusses the advantages and shortcomings of short-read and long-read sequencing technologies for accurate protein variant prediction. Your choice depends on your biological question and available resources.

### What tools are available for building sample-specific databases?

The [Spritz tool](https://pubmed.ncbi.nlm.nih.gov/32967423) provides an automated pipeline for generating protein databases annotated with sequence variations and post-translational modifications. The [Bioconductor project](https://bioconductor.org/) provides packages for genomic analysis. The [Galaxy Training Network](https://training.galaxyproject.org/) offers accessible workflow training. The [nf-core documentation](https://nf-co.re/docs) describes community pipeline standards for reproducible workflows.

### How do I control false discoveries when using sample-specific databases?

Control the false discovery rate using target-decoy approaches, where you search the data against a decoy database containing reversed or shuffled sequences and use the number of decoy hits to estimate the false discovery rate. Variant peptides should be manually inspected to confirm that the spectra support the identified sequences. Apply appropriate quality filters to variant calls during database construction to avoid including false variants.

### Can I combine standard and sample-specific databases?

Yes, many workflows use a hybrid approach where the standard database is supplemented with sample-specific sequences. This approach preserves the ability to identify canonical peptides while adding the capacity to detect variant peptides. The [Spritz paper](https://pubmed.ncbi.nlm.nih.gov/32967423) describes databases that contain both reference sequences and sample-specific variations.

### What are the limitations of sample-specific databases?

Sample-specific databases are larger than standard databases, which increases search time and may increase the false discovery rate. The database only contains variants detected in the sequencing data, so undetected variants will be absent. RNA-seq does not capture all protein sequences, and post-translational modifications require separate annotation. Results from sample-specific database searches are more complex to interpret and require validation of variant peptides.

## Related Bioinformatics Guides

- [RNA-Seq Databases: Accessing and Using Public RNA-Seq Data](/knowledge/bioinformatics/rna-seq-databases-accessing-and-using-public-rna-seq-data)
- [Single-Cell Sequencing Databases: Resources for Data Sharing and Exploration](/knowledge/bioinformatics/single-cell-sequencing-databases-resources-for-data-sharing-and-exploration)
- [Digital Pathology Guidelines: A Reference for Implementation](/knowledge/bioinformatics/digital-pathology-guidelines-a-reference-for-implementation)
- [Functional Annotation of Metagenomes: A Guide to Databases and Pipelines](/knowledge/bioinformatics/functional-annotation-of-metagenomes-a-guide-to-databases-and-pipelines)
- [Plasma Proteomics: From Sample Collection to Biomarker Discovery](/knowledge/bioinformatics/plasma-proteomics-from-sample-collection-to-biomarker-discovery)

## 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.
- [Spritz: A Proteogenomic Database Engine.](https://pubmed.ncbi.nlm.nih.gov/32967423). Journal of proteome research, 2021.
- [Accurate Prediction of Protein Sequences for Proteogenomics Data Integration.](https://pubmed.ncbi.nlm.nih.gov/34905178). Methods in molecular biology (Clifton, N.J.), 2022.
- [Longitudinal Detection of Tumor-Specific Peptides in Cerebrospinal Fluid for Pediatric Brain Tumor Surveillance.](https://doi.org/10.3390/cells15050474). 2026.
- [A multi-omics framework for survival mediation analysis of high-dimensional proteogenomic data.](https://doi.org/10.1371/journal.pcbi.1014217). 2026.
- [Deciphering RNA and protein expression discordance identifies TOP2A as a prognostic biomarker and potential therapeutic target in lung adenocarcinoma.](https://doi.org/10.1007/s12672-026-04470-x). 2026.

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