Using Proteogenomics to Discover Novel Protein-Coding Regions: From Open Reading Frames to Validated Peptides

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Proteogenomics is the analytical strategy that uses mass spectrometry data to search protein sequence databases built from genomic and transcriptomic information, instead of relying only on annotated reference proteomes. This approach allows researchers to detect peptides that originate from unannotated open reading frames, including upstream open reading frames in 5' untranslated regions, alternative reading frames within known genes, translated pseudogenes, and small open reading frames in long non-coding RNAs. The practical problem this workflow solves is straightforward: standard proteomics pipelines search against reference protein databases that contain only previously annotated proteins, so any peptide arising from a novel coding region is never matched and is silently discarded. By building custom databases from six-frame translations of the genome, filtering candidate open reading frames with transcriptomic or ribosome profiling evidence, and validating peptide-spectrum matches with stringent quality controls, researchers can discover and confirm previously unknown protein-coding regions.

This article provides a complete workflow for researchers who have access to proteomics data and want to search for novel protein-coding regions. It covers database construction, search parameters, false discovery rate control, validation strategies, and the common failure patterns that produce false positives. The workflow is designed for biology students, researchers, laboratory professionals, and life-science practitioners who need a reproducible path from raw mass spectrometry data to validated novel peptides.

Key Takeaways

• Proteogenomics uses mass spectrometry data to search protein databases built from genomic and transcriptomic information, enabling detection of peptides from unannotated open reading frames.
• Standard reference proteome databases miss novel peptides because they contain only previously annotated proteins, so any peptide from a novel coding region is never matched and silently discarded.
• Six-frame translation of the genome creates a database of all possible open reading frames, but transcriptomic and ribosome profiling evidence should filter candidates to reduce false positives.
• A two-tier search strategy first matches peptides against the reference proteome, then searches the novel database containing only unannotated open reading frames to reduce multiple testing burden.
• High-resolution mass spectrometry with parts per million mass accuracy is required for confident novel peptide identification, and sample preparation must retain small proteins to detect micropeptides.

The Problem with Reference Proteome Databases

Standard protein identification relies on matching experimental tandem mass spectra against theoretical spectra derived from a reference proteome database. A protein is identified by correlating the theoretical fragmentation spectrum derived from genomic or transcriptomic data against the experimental fragmentation mass spectrum acquired from large-scale analysis of protein mixtures. This correlation-based approach works well when the protein is present in the reference database, but it fails completely when the peptide sequence is not represented.

The human reference proteome contains approximately 20,000 annotated protein-coding genes, yet a draft map of the human proteome generated by deep proteomic profiling of 30 histologically normal human samples identified proteins encoded by 17,294 genes, accounting for roughly 84% of the total annotated protein-coding genes. The same study used a proteogenomic analysis strategy to discover novel protein-coding regions, including translated pseudogenes, non-coding RNAs, and upstream open reading frames. This finding demonstrates that even with extensive annotation efforts, the reference proteome does not capture the full set of translated products in a cell.

The gap between annotated and actual protein-coding regions is larger than most researchers expect. Proteogenomics and ribosome profiling studies have identified a large number of proteins encoded by noncoding regions in the human genome, including small open reading frames in untranslated regions of messenger RNAs and long non-coding RNAs. These small open reading frame encoded proteins are often shorter than 150 amino acids and show poor evolutionary conservation, which explains why they were missed by computational gene prediction methods that rely on conservation signals.

For researchers working with non-model organisms, the problem is even more acute. In Eucalyptus grandis, a proteogenomics study using a modified spectral correlation strategy and a dedicated de novo peptide sequencing pipeline confidently identified 41 novel peptide forms and six peptides containing at least one single amino acid substitution, with the most representative genomic class of novel peptides originating from alternative reading frames. When reference annotations are incomplete, the number of missed peptides can be substantial.

Core Principles of Proteogenomic Discovery

Six-Frame Translation as the Foundation

The starting point for proteogenomic discovery is translating the genome in all six reading frames. This produces a database of all possible open reading frames, regardless of whether they correspond to annotated genes. The resulting database is much larger than a reference proteome database, which increases search times and raises the risk of false positives. The tradeoff is that this approach makes no assumptions about which regions of the genome are protein-coding, allowing discovery of peptides from any translated locus.

The six-frame translation approach is computationally intensive but conceptually simple. Every possible open reading frame that could produce a peptide is included in the search database. This includes frames on both strands of the genome and all three reading frames per strand. The database size depends on the genome size and the minimum open reading frame length selected. Shorter minimum lengths produce larger databases but can capture small open reading frames that encode micropeptides.

Integrating Transcriptomic Evidence

A pure six-frame translation database is too large for routine searches and produces many false positives. The solution is to filter candidate open reading frames using transcriptomic evidence. RNA sequencing data provides sample-specific transcriptome information that can be used to restrict the search space to regions that are actually transcribed. This integration of next-generation sequencing technologies into the proteomics field has led to more comprehensive results and new discoveries.

RNA sequencing can capture sample-specific genomic variations, enabling focused identification of neo-antigens that arise from mutations. For novel open reading frame discovery, RNA sequencing evidence helps distinguish transcribed regions from genomic noise. The key decision is how to use the transcriptome data: assembled transcripts can be translated in all reading frames, or RNA sequencing reads can be mapped to the genome and used to define expressed regions for six-frame translation.

Ribosome Profiling for Translation Evidence

Ribosome profiling, also known as RIBO-Seq, is an RNA sequencing based technique that sequences the approximately 30 base pair fragments captured by translating ribosomes. This approach can identify translation events in small open reading frames that are considered to be non-coding, leading to the discovery of micropeptides. Ribosome profiling provides direct evidence that a region is being translated, which is stronger evidence than transcription alone.

The integration of ribosome profiling data into proteogenomic workflows addresses a specific problem: many transcribed regions are not translated, and including them in the search database increases the false discovery rate. By filtering candidate open reading frames to those with ribosome occupancy, researchers can reduce the database size while retaining regions with genuine translation evidence. This multi-omics integration is essential for identifying small translation products.

Building the Search Database

Genome Assembly and Annotation Files

The first decision is which genome assembly to use. The genome assembly defines the coordinate system for all downstream analysis, so it must be consistent across the transcriptomic and proteomic data. For human studies, the current reference assembly from NCBI provides the standard coordinate system. For non-model organisms, the quality of the genome assembly directly affects the reliability of novel peptide identifications, since assembly errors create false open reading frames.

The annotation file, typically in GTF or GFF format, provides the known gene models. This annotation is used to categorize identified peptides as annotated or novel. Peptides that map to annotated exons are expected, while peptides that map to intergenic regions, introns, or alternative reading frames of annotated genes are candidates for novel coding regions. The annotation version must be recorded, since annotation updates can reclassify novel peptides as annotated.

Custom Database Construction Workflow

Constructing a custom proteogenomic database involves several steps. First, extract the genome sequence and translate it in all six reading frames. Second, filter the resulting open reading frames by length, typically requiring a minimum of 30 to 50 amino acids for standard searches, though shorter thresholds are used when searching for micropeptides. Third, integrate transcriptomic evidence by retaining only open reading frames that overlap expressed regions. Fourth, add common contaminants and decoy sequences for false discovery rate estimation.

The database construction can be performed using command line tools and scripting. The Carpentries lessons provide foundational training in the shell, Git, and programming skills needed to manage these computational workflows. For researchers who prefer graphical interfaces, the Galaxy Training Network offers accessible workflow training and analysis tutorials that cover database construction and proteogenomic analysis steps.

Database Size and Search Time Tradeoffs

The size of the search database directly affects both search time and statistical power. Larger databases require more time to search and increase the multiple testing burden, which raises the threshold for significant peptide-spectrum matches. The practical consequence is that an unfiltered six-frame translation database can reduce sensitivity for abundant proteins while increasing the risk of false positives for novel peptides.

The solution is tiered searching. A first search against the reference proteome identifies all confidently matched peptides from annotated proteins. A second search against the novel database, containing only open reading frames not represented in the reference proteome, identifies candidate novel peptides. This two-tier approach keeps the novel database smaller and reduces the multiple testing burden for novel peptide discovery.

Mass Spectrometry Data Requirements

Instrumentation and Acquisition Modes

High-resolution mass spectrometry is essential for proteogenomic discovery. The draft map of the human proteome was generated using high-resolution Fourier-transform mass spectrometry, which provides the mass accuracy needed to distinguish novel peptides from known peptides with similar masses. Lower resolution instruments produce mass measurements that cannot reliably support novel peptide identifications.

The acquisition mode also matters. Data-dependent acquisition selects the most abundant precursor ions for fragmentation, which biases detection toward highly abundant proteins. This bias affects novel peptide discovery because small open reading frame encoded proteins are often expressed at low levels. Data-independent acquisition, which fragments all precursor ions in a systematic fashion, can improve detection of low-abundance novel peptides but produces more complex spectra that require careful analysis.

Sample Preparation Considerations

Sample preparation affects the detection of small proteins and peptides. Standard proteomics workflows often include size exclusion steps that discard proteins below a certain molecular weight, which removes exactly the small open reading frame encoded proteins that proteogenomics aims to discover. Researchers interested in micropeptides must adapt their sample preparation to retain small proteins.

The identification of small proteins and peptides has consistently proven to be challenging, and technological advances have been necessary to facilitate their identification. For micropeptide discovery, the sample preparation should avoid excessive fractionation steps that lose small proteins, and the digestion protocol should be optimized for the sequence characteristics of small open reading frame encoded proteins, which often lack the canonical trypsin cleavage sites found in larger proteins.

Data Quality Assessment Before Searching

Before searching a proteogenomic database, the quality of the mass spectrometry data should be assessed. Key metrics include the number of tandem mass spectra acquired, the precursor mass error distribution, the fragmentation quality, and the number of peptides identified in a reference proteome search. Poor quality data will produce unreliable novel peptide identifications regardless of the database construction.

The precursor mass error distribution is particularly important for proteogenomics. Novel peptides often contain single amino acid substitutions relative to annotated proteins, and these substitutions change the peptide mass. If the mass accuracy is insufficient to distinguish a substituted peptide from the annotated peptide, the identification is unreliable. High-resolution instruments with parts per million mass accuracy are required for confident novel peptide identification.

Search Strategies and Parameters

Search Engines and Their Limitations

Multiple search engines can be used for proteogenomic database searching. Each engine implements a different scoring algorithm, and the choice of engine affects the results. Using multiple search engines and requiring consensus between them can increase confidence in novel peptide identifications, but this approach also reduces sensitivity.

The search engine parameters must be set appropriately for the database type. For six-frame translation databases, the search space includes many peptides that are not present in the sample, which affects the statistical significance calculations. The false discovery rate estimation must account for the database composition, and the target-decoy approach should use a decoy database constructed in the same manner as the target database.

False Discovery Rate Control

False discovery rate control is the central quality control in proteogenomics. The standard approach is the target-decoy strategy, where the search database contains both the target sequences and reversed or shuffled decoy sequences. Peptide-spectrum matches to decoy sequences are used to estimate the false discovery rate at a given score threshold.

For novel peptide discovery, the false discovery rate must be controlled at both the peptide level and the protein level. The peptide-level false discovery rate controls the proportion of incorrectly matched spectra, while the protein-level false discovery rate controls the proportion of incorrectly identified proteins. Novel proteins identified by a single peptide require additional validation because the protein-level false discovery rate is less reliable for single-peptide identifications.

Spectrum Matching and Validation

The integrated proteogenomics analysis workflow known as IPAW combines peptide discovery, curation, and validation. This workflow includes the SpectrumAI tool for automated inspection of tandem mass spectra, which eliminates false identifications of single-residue substitution peptides. The automated spectral inspection is critical because single amino acid substitution peptides produce spectra that are nearly identical to the annotated peptide spectra, differing only in the fragment ions that contain the substituted residue.

The IPAW workflow was used to analyze proteomics data sets acquired from A431 cells and five normal human tissues using extended pH range high-resolution isoelectric focusing pre-fractionation and tandem mass tag based peptide quantitation. The results provided evidence for the translation of pseudogenes, long non-coding RNAs, short open reading frames, alternative open reading frames, N-terminal extensions, and intronic sequences. The quantitative analysis also indicated that protein production from certain pseudogenes and long non-coding RNAs is tissue specific.

Filtering Candidate Open Reading Frames

Minimum Length and Sequence Composition Filters

The minimum open reading frame length is a critical filter parameter. Standard gene prediction methods typically require open reading frames longer than 100 codons, but small open reading frame encoded proteins are often shorter than 150 amino acids and can be much shorter. The minimum length threshold must be set based on the research question: longer thresholds reduce false positives but miss micropeptides, while shorter thresholds capture small proteins but increase the database size.

Sequence composition filters can remove spurious open reading frames that arise from low complexity regions. Open reading frames with repetitive amino acid sequences or unusual amino acid composition are more likely to be false positives. These filters must be applied carefully, since some genuine small proteins have biased amino acid composition.

Transcriptomic Support Filters

Transcriptomic support is the most powerful filter for reducing the search space. RNA sequencing data from the same sample or tissue type provides evidence that a genomic region is transcribed. Open reading frames that do not overlap transcribed regions are unlikely to be translated and can be removed from the database.

The stringency of the transcriptomic filter affects the discovery rate. A permissive filter that retains any open reading frame with minimal read support will include more false candidates, while a stringent filter that requires substantial read coverage will exclude genuine novel open reading frames that are expressed at low levels. The filter threshold should be set based on the RNA sequencing depth and the expected expression levels of novel proteins.

Ribosome Occupancy Filters

Ribosome profiling data provides stronger evidence for translation than RNA sequencing alone. The presence of ribosome footprints in a region indicates active translation, and the periodicity of the footprints can reveal the reading frame. Open reading frames with ribosome occupancy in the correct reading frame are strong candidates for novel protein-coding regions.

The integration of ribosome profiling data is particularly important for small open reading frames in untranslated regions and long non-coding RNAs. These regions are transcribed but were previously considered non-coding, and ribosome profiling can identify translation events that produce micropeptides. The combination of ribosome profiling and mass spectrometry provides complementary evidence: ribosome profiling shows that translation occurs, while mass spectrometry confirms the protein product exists.

Validation of Novel Peptide Identifications

Spectral Quality Assessment

The first validation step is manual or automated inspection of the tandem mass spectra supporting novel peptide identifications. The spectrum should show a clear series of fragment ions that match the predicted fragmentation pattern of the peptide sequence. The SpectrumAI tool automates this inspection for single-residue substitution peptides, but manual inspection is still recommended for the most important novel identifications.

The spectral quality assessment should examine the coverage of the peptide sequence by fragment ions. A confident identification requires fragment ions that cover most of the peptide sequence, including the region that distinguishes the novel peptide from any annotated peptide. If the distinguishing fragment ions are absent, the identification is not reliable.

Targeted Mass Spectrometry Validation

Targeted mass spectrometry approaches provide orthogonal validation for novel peptide identifications. Parallel reaction monitoring is a targeted method that acquires high-resolution product ion spectra for specific precursor ions. This approach was used to validate novel peptide sequences in Eucalyptus grandis, providing further mass spectrometry support for the existence of the novel peptide sequences.

The targeted validation experiment should be designed to detect the specific novel peptide and, ideally, a stable isotope labeled standard of the same sequence. The retention time and fragmentation pattern of the endogenous peptide should match the standard. This validation is particularly important for novel peptides identified by a single spectrum in the discovery experiment.

Orthogonal Evidence Integration

The strongest validation comes from integrating multiple lines of evidence. A novel open reading frame is most confidently confirmed when it has ribosome profiling evidence for translation, RNA sequencing evidence for transcription, and mass spectrometry evidence for the protein product. The integration of these multi-omics data is essential for the identification of small translation products.

The review of protein-coding potential of non-canonical open reading frames in the human transcriptome discusses the approaches used for their identification, their limitations, and the reliability of these identifications. The review also discusses functionally characterized small open reading frame encoded proteins and their involvement in various biological processes and diseases, providing context for which novel identifications are most likely to be biologically meaningful.

At a Glance

Practical Implementation Steps

Step 1: Define the Research Question and Scope

The first implementation decision is the scope of the discovery effort. A focused study might search for novel open reading frames in a specific tissue or cell type, while a broader study might search across multiple samples. The scope determines the sample preparation, the number of mass spectrometry runs, and the database construction strategy.

The research question also determines which classes of novel coding regions are of interest. A study focused on micropeptides requires different sample preparation and database parameters than a study focused on translated pseudogenes or alternative reading frames. The study by the Eucalyptus grandis proteogenomics analysis identified novel peptides from alternative reading frames as the most representative genomic class, which required a database construction strategy that retained all reading frames of annotated genes.

Step 2: Acquire and Organize Input Data

The input data includes the genome assembly, the annotation file, RNA sequencing data or ribosome profiling data from the same sample type, and the mass spectrometry data. All data must be organized with clear file naming conventions and version records. The genome assembly version and annotation version must be recorded, since these define the coordinate system for all downstream analysis.

The NCBI Data Resources provide access to genome assemblies, annotation files, and sequence databases for many organisms. For researchers who need training in data management and analysis, the EMBL-EBI Training program offers learning pathways for bioinformatics data resources and practical analysis education.

Step 3: Construct the Search Database

The database construction begins with six-frame translation of the genome. The resulting open reading frames are filtered by minimum length and then by transcriptomic support. The filtered open reading frames are translated to protein sequences and combined with common contaminants and decoy sequences.

The database construction should be scripted and version controlled. The nf-core documentation provides standards for community pipelines, usage, configuration, and reproducible workflow context that can guide the development of reproducible database construction pipelines. The Bioconductor project provides official package, workflow, installation, and reproducible genomic analysis documentation that includes tools for sequence manipulation and annotation.

Step 4: Perform the Mass Spectrometry Search

The mass spectrometry data is searched against the reference proteome first, then against the novel database. The search parameters must be consistent between the two searches, with the exception of the database file. The search results are filtered to a defined false discovery rate threshold, and the novel peptide identifications are extracted.

The search should be performed with multiple search engines when possible. Consensus between search engines increases confidence, while discordant results indicate that the identification is not robust. The search parameters, including precursor mass tolerance, fragment mass tolerance, and enzyme specificity, must be recorded for reproducibility.

Step 5: Validate Novel Peptide Identifications

The candidate novel peptides are validated through spectral inspection, targeted mass spectrometry, and integration with transcriptomic and ribosome profiling evidence. The validation process should be documented, and the evidence for each novel peptide should be recorded in a structured format.

The validation process should include a clear decision criterion for which novel peptides are reported as confirmed. A peptide confirmed by multiple spectra, manual spectral inspection, and targeted mass spectrometry is a high confidence identification. A peptide identified by a single spectrum without orthogonal validation should be reported as a candidate requiring further confirmation.

Step 6: Report Results with Appropriate Confidence Levels

The reporting of novel peptide identifications must include the confidence level and the supporting evidence. The report should distinguish between high confidence identifications with multiple lines of evidence and candidate identifications that require further validation. The false discovery rate thresholds and the validation criteria should be stated explicitly.

The reporting should also include the genomic coordinates of the novel open reading frames, the predicted protein sequences, and the evidence supporting each identification. This information allows other researchers to evaluate the identifications and to design follow-up experiments.

Records and Measurements

Essential Records for Reproducibility

Reproducibility requires complete records of all analysis steps. The essential records include the genome assembly version, the annotation version, the database construction script, the search engine versions and parameters, the false discovery rate thresholds, and the validation criteria. These records should be stored with the raw data and analysis results.

Adhering to community reporting and documentation standards creates the methodological transparency necessary for independent research groups to faithfully replicate published computational analyses.

Measurements to Track During Analysis

The key measurements during analysis include the database size, the number of peptide-spectrum matches at each false discovery rate threshold, the number of novel peptides identified, and the distribution of novel peptides across genomic categories. Tracking these measurements across samples and experiments helps identify systematic issues in the workflow.

The number of novel peptides identified per sample provides a measure of the discovery rate. A sudden increase in novel peptide identifications may indicate a database construction error, while a decrease may indicate a sample preparation or mass spectrometry issue. The distribution of novel peptides across genomic categories, such as pseudogenes, long non-coding RNAs, and alternative reading frames, provides insight into the biological sources of novel translation.

Documentation of Validation Evidence

The validation evidence for each novel peptide should be documented in a structured format. The documentation should include the spectrum supporting the identification, the fragment ion coverage, the results of any targeted mass spectrometry validation, and the transcriptomic and ribosome profiling evidence for the open reading frame.

This documentation serves multiple purposes. It allows other researchers to evaluate the confidence of the identification, it provides the evidence needed for genome annotation updates, and it supports follow-up functional studies. The documentation should be stored in a format that can be shared with the research community.

Common Failure Patterns

Failure Pattern 1: Unfiltered Database Produces False Positives

The most common failure in proteogenomic discovery is searching against an unfiltered six-frame translation database. The large database increases the multiple testing burden, and the false discovery rate estimation becomes unreliable. The result is a large number of false positive novel peptide identifications that do not reproduce in validation experiments.

The solution is to filter the database using transcriptomic evidence before searching. The RNA sequencing data provides sample-specific transcriptome information that restricts the search space to transcribed regions. This filtering reduces the database size and improves the reliability of the false discovery rate estimation.

Failure Pattern 2: Single Amino Acid Substitution Peptides Misidentified

Single amino acid substitution peptides produce spectra that are nearly identical to the annotated peptide spectra. The only difference is in the fragment ions that contain the substituted residue. If these distinguishing fragment ions are not observed or are not correctly assigned, the substitution peptide is falsely identified.

The SpectrumAI tool in the IPAW workflow automates the inspection of tandem mass spectra to eliminate false identifications of single-residue substitution peptides. The automated inspection examines the fragment ions that distinguish the substitution peptide from the annotated peptide and requires that these ions are present with appropriate intensity.

Failure Pattern 3: Contaminant Peptides Misclassified as Novel

Contaminant proteins, including keratins, trypsin, and other common laboratory contaminants, can produce peptides that match novel open reading frames by chance. This is particularly problematic when the contaminant database is not included in the search, since the contaminant peptides are then matched to the closest sequence in the novel database.

The solution is to include a comprehensive contaminant database in all searches. The contaminant database should include common laboratory contaminants and the sequences of any proteins used in sample processing. Peptides matching contaminants should be excluded from novel peptide reporting.

Failure Pattern 4: Sample Preparation Loses Small Proteins

Standard proteomics sample preparation workflows often discard small proteins. The size exclusion steps, the digestion conditions, and the fractionation methods are optimized for proteins in the typical molecular weight range. Small open reading frame encoded proteins, which are often shorter than 150 amino acids, are lost during these steps.

The solution is to adapt the sample preparation for small protein retention. This may require different fractionation methods, modified digestion conditions, and mass spectrometry acquisition parameters optimized for small peptides. The identification of small proteins and peptides has consistently proven to be challenging, and the sample preparation must be designed with this challenge in mind.

Failure Pattern 5: Annotation Updates Invalidate Novel Identifications

A novel peptide identified in one annotation version may become an annotated peptide in a later annotation version. This is not a failure of the identification but a consequence of annotation improvement. The reverse situation is more problematic: a peptide annotated in one version may be removed in a later version, which affects the interpretation of the identification.

The solution is to record the annotation version used in the analysis and to re-evaluate novel identifications when the annotation is updated. The NCBI Data Resources provide access to annotation updates and the tools to compare annotations across versions.

Limitations and Interpretation Boundaries

Mass Spectrometry Detection Limits

Mass spectrometry cannot detect all proteins in a sample. The detection limit depends on the protein abundance, the sample complexity, the mass spectrometry instrument, and the acquisition method. Novel proteins expressed at low levels may not be detected, and the absence of mass spectrometry evidence does not prove that a region is not translated.

The draft map of the human proteome identified proteins encoded by 17,294 genes, accounting for approximately 84% of the total annotated protein-coding genes. The remaining 16% were not detected, either because they were not expressed in the sampled tissues or because they were below the detection limit. This limitation applies equally to novel protein discovery.

Database Completeness and Errors

The genome assembly and annotation files used to construct the search database may contain errors. Assembly errors create false open reading frames, while annotation errors misclassify peptides as novel or annotated. The quality of the genome assembly directly affects the reliability of novel peptide identifications.

For non-model organisms, the genome assembly quality is often lower than for model organisms. The Eucalyptus grandis study used a modified spectral correlation strategy and a dedicated de novo peptide sequencing pipeline to address the challenges of working with a less complete genome annotation. Researchers working with non-model organisms should assess the genome assembly quality before interpreting novel peptide identifications.

Evolutionary Conservation and Functional Significance

Small open reading frame encoded proteins often show poor evolutionary conservation. This lack of conservation makes computational prediction difficult but does not indicate a lack of function. A subset of small open reading frame encoded proteins have been functionally characterized and shown to play important roles in fundamental biological processes including cardiac and muscle function, DNA repair, embryonic development, and various human diseases.

The lack of evolutionary conservation also complicates validation. Orthology-based arguments cannot be used to support the identification, and the functional significance must be established through experimental studies. The review of protein-coding potential of non-canonical open reading frames discusses the distinctive features of small open reading frame encoded proteins compared to canonical proteins and the challenges associated with annotating these in protein reference databases.

Tissue Specificity of Novel Protein Expression

The expression of novel proteins can be tissue specific. The IPAW analysis indicated that protein production from certain pseudogenes and long non-coding RNAs is tissue specific. This tissue specificity means that a novel protein detected in one tissue may not be detected in another tissue, and the absence of detection in a particular tissue does not rule out translation of the open reading frame.

The tissue specificity also affects the design of validation experiments. Targeted mass spectrometry validation should use the same tissue or cell type as the discovery experiment. If the novel protein is expressed at low levels or in a restricted set of tissues, the validation experiment must be designed with sufficient sensitivity.

Safety and Data Management Context

Data Sharing and Reproducibility Standards

Proteogenomic data should be shared according to community standards. The mass spectrometry data should be deposited in public repositories, and the analysis scripts and parameters should be made available. The Eucalyptus grandis study made data available via ProteomeXchange with identifier PXD022110, following the community standard for proteomics data sharing.

Researchers can leverage the Galaxy Training Network's open infrastructure for transparent, step-by-step analysis courses covering the full NGS toolkit.

Computational Resource Requirements

Proteogenomic analysis requires substantial computational resources. The six-frame translation of a genome produces a large database, and searching mass spectrometry data against this database requires significant processing time. The computational requirements depend on the genome size, the number of mass spectrometry runs, and the search parameters.

Researchers should plan for the computational requirements before starting the analysis. The Carpentries lessons provide foundational training in the shell, Git, and programming skills needed to manage these computational workflows. The Bioconductor project provides official package, workflow, installation, and reproducible genomic analysis documentation that includes tools for sequence manipulation and annotation.

Professional Escalation Criteria

Researchers should escalate to more specialized expertise when certain conditions are encountered. If the false discovery rate cannot be controlled at the desired threshold, if the validation experiments consistently fail to confirm novel peptide identifications, or if the results are challenged by reviewers or collaborators, consultation with a proteomics bioinformatics specialist is recommended.

The EMBL-EBI Training program offers learning pathways for bioinformatics data resources and practical analysis education. The Galaxy Training Network provides accessible workflow training and analysis tutorials. These resources can help researchers build the skills needed to troubleshoot proteogenomic analysis problems.

Frequently Asked Questions

What is the minimum open reading frame length for proteogenomic discovery?

The minimum open reading frame length depends on the research question. Standard gene prediction methods typically require open reading frames longer than 100 codons, but small open reading frame encoded proteins are often shorter than 150 amino acids and can be much shorter. For micropeptide discovery, the minimum length threshold should be set lower, often around 30 to 50 amino acids, to capture small proteins. The tradeoff is that shorter thresholds produce larger databases and increase the false discovery rate. The threshold should be set based on the specific classes of novel coding regions of interest and validated with known positive controls.

How do I distinguish a novel peptide from a single amino acid variant of an annotated peptide?

A single amino acid substitution peptide produces a spectrum that is nearly identical to the annotated peptide spectrum, differing only in the fragment ions that contain the substituted residue. The distinguishing fragment ions must be observed with appropriate intensity to support the substitution. The SpectrumAI tool in the integrated proteogenomics analysis workflow automates this inspection and eliminates false identifications of single-residue substitution peptides. Manual spectral inspection is recommended for the most important identifications, and targeted mass spectrometry validation provides orthogonal confirmation.

What false discovery rate threshold should I use for novel peptide identification?

The false discovery rate threshold should be more stringent for novel peptide identifications than for standard protein identifications. A common approach is to require a 1% peptide level false discovery rate for the overall search and a more stringent threshold, such as 0.1%, for novel peptides that are reported as high confidence identifications. The false discovery rate estimation must use a decoy database constructed in the same manner as the target database. The threshold should be stated explicitly in the reporting of results.

Why do my novel peptide identifications fail to validate by targeted mass spectrometry?

Validation failure can result from several causes. The original identification may be a false positive, particularly if the spectrum quality was marginal or the false discovery rate control was inadequate. The novel protein may be expressed at levels below the detection limit of the targeted assay. The sample used for validation may differ from the discovery sample in tissue type, developmental stage, or treatment conditions. The targeted assay may not be optimized for the peptide sequence, particularly if the peptide has unusual fragmentation properties. Each of these possibilities should be investigated systematically.

Can I use RNA sequencing data alone to predict novel protein-coding regions?

RNA sequencing data provides evidence of transcription but not translation. Many transcribed regions are not translated, and the presence of RNA does not confirm the existence of a protein product. Ribosome profiling provides stronger evidence for translation, since it captures the fragments protected by translating ribosomes. Mass spectrometry provides the most direct evidence for the protein product. The strongest identifications integrate all three lines of evidence. RNA sequencing data alone can identify candidate open reading frames, but these candidates require translation evidence for confirmation.

How do I handle novel peptides that map to multiple genomic locations?

Peptides that map to multiple genomic locations are ambiguous and should be handled with caution. If the peptide sequence is present in multiple open reading frames, the mass spectrometry evidence cannot distinguish which locus produced the peptide. The peptide should be reported as mapping to multiple locations, and the confidence in any single locus should be reduced. If the peptide is unique to one locus at the nucleotide level but the amino acid sequence is shared, the genomic assignment is more confident. The mapping should be performed at both the nucleotide and amino acid levels.

What evidence is needed to update genome annotations with novel protein-coding regions?

Genome annotation updates require strong evidence for the novel protein-coding region. The minimum evidence is a confident mass spectrometry identification with a spectrum that clearly supports the peptide sequence. Stronger evidence includes multiple independent spectra, targeted mass spectrometry validation, ribosome profiling evidence for translation, and RNA sequencing evidence for transcription. The evidence should be submitted to the genome annotation resources, such as those provided by NCBI, following their submission guidelines. The annotation update process evaluates the evidence and may require additional validation.

How do I report novel peptide identifications in a publication?

Novel peptide identifications should be reported with the peptide sequence, the genomic coordinates of the open reading frame, the confidence level, and the supporting evidence. The false discovery rate thresholds and validation criteria should be stated explicitly. The mass spectrometry data should be deposited in a public repository, and the analysis scripts and parameters should be made available. The reporting should distinguish between high confidence identifications with multiple lines of evidence and candidate identifications that require further validation. The limitations of the identification approach should be discussed, including the detection limits and the potential for false positives.

Related Bioinformatics Guides

• Genomic Data Analysis Tools: A Comparative Guide for Researchers
• Olink Proteomics: A Practical Guide to Panel Selection and Data Interpretation
• Proteomics Mass Spectrometry: From Sample Preparation to Data Analysis
• TMT Proteomics: Experimental Design, Labeling, and Data Analysis
• Spatial Proteomics Platforms: A Comparison of Commercial and Open-Source Solutions

References and Further Reading

• NCBI Data Resources. National Center for Biotechnology Information.
• EMBL-EBI Training. European Bioinformatics Institute.
• Bioconductor. Bioconductor Project.
• Galaxy Training Network. Galaxy Project.
• nf-core Documentation. nf-core.
• The Carpentries Lessons. The Carpentries.
• A draft map of the human proteome.. Nature, 2014.
• Identification of Small Novel Coding Sequences, a Proteogenomics Endeavor.. Advances in experimental medicine and biology, 2016.
• Identification of novel protein-coding sequences in Eucalyptus grandis plants by high-resolution mass spectrometry.. Biochimica et biophysica acta. Proteins and proteomics, 2021.
• Protein-coding potential of non-canonical open reading frames in human transcriptome.. Biochemical and biophysical research communications, 2023.
• Discovery of coding regions in the human genome by integrated proteogenomics analysis workflow.. Nature communications, 2018.

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